diff --git a/addons/hr/models/hr.py b/addons/hr/models/hr.py
index 89b883e6df4..cccb266981b 100644
--- a/addons/hr/models/hr.py
+++ b/addons/hr/models/hr.py
@@ -106,7 +106,7 @@ class Employee(models.Model):
     @api.model
     def _default_image(self):
         image_path = get_module_resource('hr', 'static/src/img', 'default_image.png')
-        return tools.image_resize_image_big(base64.b64encode(open(image_path, 'rb').read()))
+        return tools.image_process(base64.b64encode(open(image_path, 'rb').read()), size=tools.IMAGE_BIG_SIZE)
 
     # resource and user
     # required on the resource, make sure required="True" set in the view
diff --git a/addons/im_livechat/models/im_livechat_channel.py b/addons/im_livechat/models/im_livechat_channel.py
index 6a568bbb269..9ef4df884f7 100644
--- a/addons/im_livechat/models/im_livechat_channel.py
+++ b/addons/im_livechat/models/im_livechat_channel.py
@@ -21,7 +21,7 @@ class ImLivechatChannel(models.Model):
 
     def _default_image(self):
         image_path = modules.get_module_resource('im_livechat', 'static/src/img', 'default.png')
-        return tools.image_resize_image_big(base64.b64encode(open(image_path, 'rb').read()))
+        return tools.image_process(base64.b64encode(open(image_path, 'rb').read()), size=tools.IMAGE_BIG_SIZE)
 
     def _default_user_ids(self):
         return [(6, 0, [self._uid])]
diff --git a/addons/mail/models/mail_channel.py b/addons/mail/models/mail_channel.py
index 1ae85a249ca..43a3dd54703 100644
--- a/addons/mail/models/mail_channel.py
+++ b/addons/mail/models/mail_channel.py
@@ -64,7 +64,7 @@ class Channel(models.Model):
 
     def _get_default_image(self):
         image_path = modules.get_module_resource('mail', 'static/src/img', 'groupdefault.png')
-        return tools.image_resize_image_big(base64.b64encode(open(image_path, 'rb').read()))
+        return tools.image_process(base64.b64encode(open(image_path, 'rb').read()), size=tools.IMAGE_BIG_SIZE)
 
     @api.model
     def default_get(self, fields):
diff --git a/addons/payment/models/payment_acquirer.py b/addons/payment/models/payment_acquirer.py
index 8fa4d67c36a..415be96bd27 100644
--- a/addons/payment/models/payment_acquirer.py
+++ b/addons/payment/models/payment_acquirer.py
@@ -7,11 +7,10 @@ from datetime import datetime
 from dateutil import relativedelta
 import pprint
 
-from odoo import api, exceptions, fields, models, _
-from odoo.tools import consteq, float_round, image_resize_images, image_resize_image, ustr
+from odoo import api, exceptions, fields, models, _, SUPERUSER_ID
+from odoo.tools import consteq, float_round, image_resize_images, image_process, ustr
 from odoo.addons.base.models import ir_module
 from odoo.exceptions import ValidationError
-from odoo import api, SUPERUSER_ID
 from odoo.tools.misc import DEFAULT_SERVER_DATETIME_FORMAT
 from odoo.tools.misc import formatLang
 
@@ -525,16 +524,16 @@ class PaymentIcon(models.Model):
         for vals in vals_list:
             if 'image' in vals:
                 image = ustr(vals['image'] or '').encode('utf-8')
-                vals['image_payment_form'] = image_resize_image(image, size=(45,30))
-                vals['image'] = image_resize_image(image, size=(64,64))
+                vals['image_payment_form'] = image_process(image, size=(45,30))
+                vals['image'] = image_process(image, size=(64,64))
         return super(PaymentIcon, self).create(vals_list)
 
     @api.multi
     def write(self, vals):
         if 'image' in vals:
             image = ustr(vals['image'] or '').encode('utf-8')
-            vals['image_payment_form'] = image_resize_image(image, size=(45,30))
-            vals['image'] = image_resize_image(image, size=(64,64))
+            vals['image_payment_form'] = image_process(image, size=(45,30))
+            vals['image'] = image_process(image, size=(64,64))
         return super(PaymentIcon, self).write(vals)
 
 class PaymentTransaction(models.Model):
diff --git a/addons/product/models/product.py b/addons/product/models/product.py
index 99a314662dd..4d566f2b100 100644
--- a/addons/product/models/product.py
+++ b/addons/product/models/product.py
@@ -168,7 +168,7 @@ class ProductProduct(models.Model):
         for record in self:
             images = tools.image_get_resized_images(record.image_raw_original, big_name=False)
             record.image_raw_big = tools.image_get_resized_images(record.image_raw_original,
-                large_name=False, medium_name=False, small_name=False, preserve_aspect_ratio=True)['image']
+                large_name=False, medium_name=False, small_name=False)['image']
             record.image_raw_large = images['image_large']
             record.image_raw_medium = images['image_medium']
             record.image_raw_small = images['image_small']
diff --git a/addons/purchase/controllers/portal.py b/addons/purchase/controllers/portal.py
index 8e5cea717ac..171373bbec5 100644
--- a/addons/purchase/controllers/portal.py
+++ b/addons/purchase/controllers/portal.py
@@ -7,7 +7,7 @@ from collections import OrderedDict
 from odoo import http
 from odoo.exceptions import AccessError, MissingError
 from odoo.http import request
-from odoo.tools import image_resize_image
+from odoo.tools import image_process
 from odoo.tools.translate import _
 from odoo.addons.portal.controllers.portal import pager as portal_pager, CustomerPortal
 from odoo.addons.web.controllers.main import Binary
@@ -28,7 +28,7 @@ class CustomerPortal(CustomerPortal):
         def resize_to_48(b64source):
             if not b64source:
                 b64source = base64.b64encode(Binary().placeholder())
-            return image_resize_image(b64source, size=(48, 48))
+            return image_process(b64source, size=(48, 48))
 
         values = {
             'order': order,
diff --git a/addons/web/controllers/main.py b/addons/web/controllers/main.py
index cf33f0232a4..a14a02ebb9b 100644
--- a/addons/web/controllers/main.py
+++ b/addons/web/controllers/main.py
@@ -36,7 +36,7 @@ import odoo
 import odoo.modules.registry
 from odoo.api import call_kw, Environment
 from odoo.modules import get_module_path, get_resource_path
-from odoo.tools import limited_image_resize, topological_sort, html_escape, pycompat
+from odoo.tools import image_process, topological_sort, html_escape, pycompat
 from odoo.tools.mimetypes import guess_mimetype
 from odoo.tools.translate import _
 from odoo.tools.misc import str2bool, xlwt, file_open
@@ -971,11 +971,6 @@ class Binary(http.Controller):
         addons_path = http.addons_manifest['web']['addons_path']
         return open(os.path.join(addons_path, 'web', 'static', 'src', 'img', image), 'rb').read()
 
-    def force_contenttype(self, headers, contenttype='image/png'):
-        dictheaders = dict(headers)
-        dictheaders['Content-Type'] = contenttype
-        return list(dictheaders.items())
-
     @http.route(['/web/content',
         '/web/content/<string:xmlid>',
         '/web/content/<string:xmlid>/<string:filename>',
@@ -1023,30 +1018,25 @@ class Binary(http.Controller):
         '/web/image/<int:id>-<string:unique>/<int:width>x<int:height>/<string:filename>'], type='http', auth="public")
     def content_image(self, xmlid=None, model='ir.attachment', id=None, field='datas',
                       filename_field='datas_fname', unique=None, filename=None, mimetype=None,
-                      download=None, width=0, height=0, crop=False, access_token=None, avoid_if_small=False,
-                      upper_limit=False, placeholder='placeholder.png', **kw):
-        status, headers, content = request.env['ir.http'].binary_content(
+                      download=None, width=0, height=0, crop=False, access_token=None,
+                      placeholder='placeholder.png', **kwargs):
+        status, headers, image_base64 = request.env['ir.http'].binary_content(
             xmlid=xmlid, model=model, id=id, field=field, unique=unique, filename=filename,
             filename_field=filename_field, download=download, mimetype=mimetype,
             default_mimetype='image/png', access_token=access_token)
 
         if status == 301 or (status != 200 and download):
-            return request.env['ir.http']._response_by_status(status, headers, content)
-        if not content:
-            content = base64.b64encode(self.placeholder(image=placeholder))
-            headers = self.force_contenttype(headers, contenttype='image/png')
+            return request.env['ir.http']._response_by_status(status, headers, image_base64)
+        if not image_base64:
+            image_base64 = base64.b64encode(self.placeholder(image=placeholder))
             if not (width or height):
-                suffix = 'big' if field == 'image' else field.split('_')[-1]
-                if suffix in ('small', 'medium', 'large', 'big'):
-                    content = getattr(odoo.tools, 'image_resize_image_%s' % suffix)(content)
+                width, height = odoo.tools.image_guess_size_from_field_name(field)
 
+        image_base64 = image_process(image_base64, (width, height), crop=crop)
 
-        content = limited_image_resize(
-            content, width=width, height=height, crop=crop, upper_limit=upper_limit, avoid_if_small=avoid_if_small)
-
-        image_base64 = base64.b64decode(content)
-        headers.append(('Content-Length', len(image_base64)))
-        response = request.make_response(image_base64, headers)
+        content = base64.b64decode(image_base64)
+        headers = http.set_safe_image_headers(headers, content)
+        response = request.make_response(content, headers)
         response.status_code = status
         return response
 
diff --git a/addons/web/models/ir_qweb.py b/addons/web/models/ir_qweb.py
index d1d792e8f54..e8a966647e3 100644
--- a/addons/web/models/ir_qweb.py
+++ b/addons/web/models/ir_qweb.py
@@ -45,7 +45,6 @@ class Image(models.AbstractModel):
 
         sha = hashlib.sha1(str(getattr(record, '__last_update')).encode('utf-8')).hexdigest()[0:7]
         max_size = '' if max_size is None else '/%s' % max_size
-        avoid_if_small = '&avoid_if_small=true' if options.get('avoid_if_small') else ''
 
         if options.get('filename-field') and getattr(record, options['filename-field'], None):
             filename = record[options['filename-field']]
@@ -54,7 +53,7 @@ class Image(models.AbstractModel):
         else:
             filename = record.display_name
 
-        src = '/web/image/%s/%s/%s%s/%s?unique=%s%s' % (record._name, record.id, options.get('preview_image', field_name), max_size, url_quote(filename), sha, avoid_if_small)
+        src = '/web/image/%s/%s/%s%s/%s?unique=%s' % (record._name, record.id, options.get('preview_image', field_name), max_size, url_quote(filename), sha)
 
         if options.get('alt-field') and getattr(record, options['alt-field'], None):
             alt = escape(record[options['alt-field']])
diff --git a/addons/web/static/src/js/views/kanban/kanban_record.js b/addons/web/static/src/js/views/kanban/kanban_record.js
index 2794d871e67..70afa09c066 100644
--- a/addons/web/static/src/js/views/kanban/kanban_record.js
+++ b/addons/web/static/src/js/views/kanban/kanban_record.js
@@ -177,21 +177,19 @@ var KanbanRecord = Widget.extend({
      * @param {string} model the name of the model
      * @param {string} field the name of the field
      * @param {integer} id the id of the resource
-     * @param {integer} cache the cache duration, in seconds
-     * @param {Object} options
+     * @param {string} placeholder
      * @returns {string} the url of the image
      */
-    _getImageURL: function (model, field, id, cache, options) {
-        options = options || {};
+    _getImageURL: function (model, field, id, placeholder) {
         id = (_.isArray(id) ? id[0] : id) || false;
-        field = options.preview_image || field;
+        placeholder = placeholder || "/web/static/src/img/placeholder.png";
         var isCurrentRecord = this.modelName === model && this.recordData.id === id;
         var url;
         if (isCurrentRecord && this.record[field] && this.record[field].raw_value && !utils.is_bin_size(this.record[field].raw_value)) {
             // Use magic-word technique for detecting image type
             url = 'data:image/' + this.file_type_magic_word[this.record[field].raw_value[0]] + ';base64,' + this.record[field].raw_value;
         } else if (!model || !field || !id || (isCurrentRecord && this.record[field] && !this.record[field].raw_value)) {
-            url = "/web/static/src/img/placeholder.png";
+            url = placeholder;
         } else {
             var session = this.getSession();
             var params = {
@@ -202,10 +200,6 @@ var KanbanRecord = Widget.extend({
             if (isCurrentRecord) {
                 params.unique = this.record.__last_update && this.record.__last_update.value.replace(/[^0-9]/g, '');
             }
-            if (cache !== undefined) {
-                // Set the cache duration in seconds.
-                params.cache = parseInt(cache, 10);
-            }
             url = session.url('/web/image', params);
         }
         return url;
diff --git a/addons/web/static/src/scss/kanban_view.scss b/addons/web/static/src/scss/kanban_view.scss
index e2c550f0a42..7ab56e2bbb3 100644
--- a/addons/web/static/src/scss/kanban_view.scss
+++ b/addons/web/static/src/scss/kanban_view.scss
@@ -146,7 +146,15 @@
                 margin-right: $o-kanban-inside-hgutter*2;
                 flex: 1 0 $o-kanban-image-fill-width;
                 min-height: 95px;
-                background: {size: cover; position: center };
+                background: {
+                    size: cover;
+                    position: center;
+                    repeat: no-repeat;
+                };
+
+                &.o_kanban_image_full {
+                    background-size: contain;
+                }
             }
 
             // Adapt default 'o_kanban_image' element if present.
@@ -158,6 +166,10 @@
                 flex: 0 0 $o-kanban-image-width;
                 min-height: $o-kanban-image-width;
                 align-self: baseline;
+                background: {
+                    repeat: no-repeat;
+                    position: center;
+                }
 
                 // On medium screen size, align 'o_kanban_image' to 'o_kanban_image_fill_left' elements.
                 @include media-breakpoint-up(md) {
diff --git a/addons/web/tests/test_image.py b/addons/web/tests/test_image.py
index e17073d0fb7..1e4b0c3c5b5 100644
--- a/addons/web/tests/test_image.py
+++ b/addons/web/tests/test_image.py
@@ -13,10 +13,10 @@ class TestImage(HttpCase):
         """The goal of this test is to make sure the placeholder image is
         resized appropriately depending on the given URL parameters."""
 
-        # CASE: resize placeholder to given size
+        # CASE: resize placeholder, given size but original ratio is always kept
         response = self.url_open('/web/image/0/200x150')
         image = Image.open(io.BytesIO(response.content))
-        self.assertEqual(image.size, (200, 150))
+        self.assertEqual(image.size, (150, 150))
 
         # CASE: resize placeholder to small
         response = self.url_open('/web/image/fake/0/image_small')
diff --git a/addons/web_editor/controllers/main.py b/addons/web_editor/controllers/main.py
index 8360161fb3d..edc23e0da76 100644
--- a/addons/web_editor/controllers/main.py
+++ b/addons/web_editor/controllers/main.py
@@ -158,27 +158,17 @@ class Web_Editor(http.Controller):
     #------------------------------------------------------
     @http.route('/web_editor/add_image_base64', type='json', auth='user', methods=['POST'], website=True)
     def add_image_base64(self, res_model, res_id, image_base64, filename, disable_optimization=None, **kwargs):
-        data = base64.b64decode(image_base64)
-        attachment = self._image_to_attachment(res_model, res_id, data, filename, filename, disable_optimization=disable_optimization)
+        attachment = self._image_to_attachment(res_model, res_id, image_base64, filename, filename, disable_optimization=disable_optimization)
         return attachment.read(['name', 'mimetype', 'checksum', 'url', 'res_id', 'res_model', 'access_token'])[0]
 
-    def _image_to_attachment(self, res_model, res_id, data, name, datas_fname, disable_optimization=None):
+    def _image_to_attachment(self, res_model, res_id, image_base64, name, datas_fname, disable_optimization=None):
         Attachments = request.env['ir.attachment']
-        try:
-            image = Image.open(io.BytesIO(data))
-            w, h = image.size
-            if w*h > 42e6: # Nokia Lumia 1020 photo resolution
-                raise ValueError(
-                    u"Image size excessive, uploaded images must be smaller "
-                    u"than 42 million pixel")
-            if not disable_optimization and image.format in ('PNG', 'JPEG'):
-                data = tools.image_save_for_web(image)
-        except IOError:
-            pass
+        if not disable_optimization:
+            image_base64 = tools.image_process(image_base64, verify_resolution=True)
         attachment = Attachments.create({
             'name': name,
             'datas_fname': datas_fname,
-            'datas': base64.b64encode(data),
+            'datas': image_base64,
             'public': res_model == 'ir.ui.view',
             'res_id': res_id,
             'res_model': res_model,
@@ -224,12 +214,12 @@ class Web_Editor(http.Controller):
             try:
                 attachments = request.env['ir.attachment']
                 for c_file in request.httprequest.files.getlist('upload'):
-                    data = c_file.read()
+                    image_base64 = base64.b64encode(c_file.read())
                     name = c_file.filename
                     datas_fname = name
                     if filters:
                         datas_fname = filters + '_' + datas_fname
-                    attachments += self._image_to_attachment(res_model, res_id, data, name, datas_fname, disable_optimization=disable_optimization)
+                    attachments += self._image_to_attachment(res_model, res_id, image_base64, name, datas_fname, disable_optimization=disable_optimization)
                 uploads += attachments.read(['name', 'mimetype', 'checksum', 'url', 'res_id', 'res_model', 'access_token'])
             except Exception as e:
                 logger.exception("Failed to upload image to attachment")
diff --git a/addons/web_unsplash/controllers/main.py b/addons/web_unsplash/controllers/main.py
index 219342d1c39..ddebf07e327 100644
--- a/addons/web_unsplash/controllers/main.py
+++ b/addons/web_unsplash/controllers/main.py
@@ -1,14 +1,14 @@
 # -*- coding: utf-8 -*-
 # Part of Odoo. See LICENSE file for full copyright and licensing details.
 import base64
-import io
 import logging
+import mimetypes
 import requests
 import werkzeug.utils
 
-from PIL import Image
 from odoo import http, tools, _
 from odoo.http import request
+from odoo.tools.mimetypes import guess_mimetype
 from werkzeug.urls import url_encode
 
 logger = logging.getLogger(__name__)
@@ -89,8 +89,7 @@ class Web_Unsplash(http.Controller):
                     continue
 
                 # get mime-type of image url because unsplash url dosn't contains mime-types in url
-                mimetype = req.headers.get('Content-Type')
-                datas = req.content
+                image_base64 = base64.b64encode(req.content)
             except requests.exceptions.ConnectionError as e:
                 logger.exception("Connection Error: " + str(e))
                 continue
@@ -98,13 +97,10 @@ class Web_Unsplash(http.Controller):
                 logger.exception("Timeout: " + str(e))
                 continue
 
-            # optimized image before save
-            if mimetype in ('image/jpeg', 'image/png'):
-                image = Image.open(io.BytesIO(datas))
-                if image.format in ('PNG', 'JPEG'):
-                    datas = tools.image_save_for_web(image)
-                    # append image extension in name
-                    query += '.' + str.lower(image.format)
+            image_base64 = tools.image_process(image_base64)
+            mimetype = guess_mimetype(base64.b64decode(image_base64))
+            # append image extension in name
+            query += mimetypes.guess_extension(mimetype) or ''
 
             # /unsplash/5gR788gfd/lion
             url_frags = ['unsplash', key, query]
@@ -114,7 +110,7 @@ class Web_Unsplash(http.Controller):
                 'url': '/' + '/'.join(url_frags),
                 'datas_fname': '_'.join(url_frags),
                 'mimetype': mimetype,
-                'datas': base64.b64encode(datas),
+                'datas': image_base64,
                 'public': res_model == 'ir.ui.view',
                 'res_id': res_id,
                 'res_model': res_model,
diff --git a/addons/website/data/website_data.xml b/addons/website/data/website_data.xml
index 9701610dde3..737ff819de8 100644
--- a/addons/website/data/website_data.xml
+++ b/addons/website/data/website_data.xml
@@ -156,7 +156,6 @@
             <field name="domain"></field>
             <field name="company_id" ref="base.main_company"/>
             <field name="user_id" ref="base.public_user"/>
-            <field name="favicon" type="base64" file="web/static/src/img/favicon.ico"/>
             <field name="homepage_id" ref="homepage_page"/>
         </record>
 
diff --git a/addons/website/models/website.py b/addons/website/models/website.py
index 0261b1fc616..4f55ed27298 100644
--- a/addons/website/models/website.py
+++ b/addons/website/models/website.py
@@ -1,6 +1,7 @@
 # -*- coding: utf-8 -*-
 # Part of Odoo. See LICENSE file for full copyright and licensing details.
 
+import base64
 import inspect
 import logging
 import hashlib
@@ -14,6 +15,7 @@ from odoo.addons.http_routing.models.ir_http import slugify, _guess_mimetype
 from odoo.addons.website.models.ir_http import sitemap_qs2dom
 from odoo.addons.portal.controllers.portal import pager
 from odoo.http import request
+from odoo.modules.module import get_resource_path
 from odoo.osv.expression import FALSE_DOMAIN
 from odoo.tools.translate import _
 
@@ -100,7 +102,13 @@ class Website(models.Model):
     partner_id = fields.Many2one(related='user_id.partner_id', relation='res.partner', string='Public Partner', readonly=False)
     menu_id = fields.Many2one('website.menu', compute='_compute_menu', string='Main Menu')
     homepage_id = fields.Many2one('website.page', string='Homepage')
-    favicon = fields.Binary(string="Website Favicon", help="This field holds the image used to display a favicon on the website.")
+
+    def _default_favicon(self):
+        img_path = get_resource_path('web', 'static/src/img/favicon.ico')
+        with tools.file_open(img_path, 'rb') as f:
+            return base64.b64encode(f.read())
+
+    favicon = fields.Binary(string="Website Favicon", help="This field holds the image used to display a favicon on the website.", default=_default_favicon)
     theme_id = fields.Many2one('ir.module.module', help='Installed theme')
 
     specific_user_account = fields.Boolean('Specific User Account', help='If True, new accounts will be associated to the current website')
@@ -122,6 +130,8 @@ class Website(models.Model):
 
     @api.model
     def create(self, vals):
+        self._handle_favicon(vals)
+
         if 'user_id' not in vals:
             company = self.env['res.company'].browse(vals.get('company_id'))
             vals['user_id'] = company._get_public_user().id if company else self.env.ref('base.public_user').id
@@ -138,6 +148,8 @@ class Website(models.Model):
 
     @api.multi
     def write(self, values):
+        self._handle_favicon(values)
+
         self._get_languages.clear_cache(self)
         if 'company_id' in values and 'user_id' not in values:
             company = self.env['res.company'].browse(values['company_id'])
@@ -149,6 +161,11 @@ class Website(models.Model):
             self.env['ir.qweb'].clear_caches()
         return result
 
+    @api.model
+    def _handle_favicon(self, vals):
+        if 'favicon' in vals:
+            vals['favicon'] = tools.image_process(vals['favicon'], size=(256, 256), crop='center', output_format='ICO')
+
     # ----------------------------------------------------------
     # Page Management
     # ----------------------------------------------------------
diff --git a/addons/website/tests/__init__.py b/addons/website/tests/__init__.py
index 85ecad1cba8..f1f4a459cee 100644
--- a/addons/website/tests/__init__.py
+++ b/addons/website/tests/__init__.py
@@ -9,4 +9,5 @@ from . import test_ui
 from . import test_views
 from . import test_menu
 from . import test_page
+from . import test_website_favicon
 from . import test_website_reset_password
diff --git a/addons/website/tests/test_website_favicon.py b/addons/website/tests/test_website_favicon.py
new file mode 100644
index 00000000000..3346ccd459f
--- /dev/null
+++ b/addons/website/tests/test_website_favicon.py
@@ -0,0 +1,36 @@
+# -*- coding: utf-8 -*-
+# Part of Odoo. See LICENSE file for full copyright and licensing details.
+
+from PIL import Image
+
+from odoo.tests import tagged
+from odoo.tests.common import TransactionCase
+from odoo.tools import base64_to_image, image_to_base64
+
+
+@tagged('post_install', '-at_install')
+class TestWebsiteResetPassword(TransactionCase):
+
+    def test_01_website_favicon(self):
+        """The goal of this test is to make sure the favicon is correctly
+        handled on the website."""
+
+        # Test setting an Ico file directly, done through create
+        Website = self.env['website']
+
+        website = Website.create({
+            'name': 'Test Website',
+            'favicon': Website._default_favicon(),
+        })
+
+        image = base64_to_image(website.favicon)
+        self.assertEqual(image.format, 'ICO')
+
+        # Test setting a JPEG file that is too big, done through write
+        bg_color = (135, 90, 123)
+        image = Image.new('RGB', (1920, 1080), color=bg_color)
+        website.favicon = image_to_base64(image, 'JPEG')
+        image = base64_to_image(website.favicon)
+        self.assertEqual(image.format, 'ICO')
+        self.assertEqual(image.size, (256, 256))
+        self.assertEqual(image.getpixel((0, 0)), bg_color)
diff --git a/addons/website_partner/views/website_partner_templates.xml b/addons/website_partner/views/website_partner_templates.xml
index c8682008619..2f5c58b1dee 100644
--- a/addons/website_partner/views/website_partner_templates.xml
+++ b/addons/website_partner/views/website_partner_templates.xml
@@ -17,7 +17,7 @@
 <template id="partner_detail" name="Partner Details">
     <h1 class="col-lg-12 text-center" id="partner_name" t-field="partner.display_name"/>
     <div class="col-lg-4">
-        <div t-field="partner.image" t-options='{"widget": "image", "class": "d-block mx-auto mb16", "max_width": 512, "avoid_if_small": True}'/>
+        <div t-field="partner.image" t-options='{"widget": "image", "class": "d-block mx-auto mb16", "max_width": 512}'/>
         <address>
              <div t-field="partner.self" t-options='{
                  "widget": "contact",
diff --git a/addons/website_profile/controllers/main.py b/addons/website_profile/controllers/main.py
index 9c6d31c2efc..5d379643e5d 100644
--- a/addons/website_profile/controllers/main.py
+++ b/addons/website_profile/controllers/main.py
@@ -32,18 +32,10 @@ class WebsiteProfile(http.Controller):
             return user.website_published and user.karma > 0
         return False
 
-    def _get_default_avatar(self, field, headers, width, height):
+    def _get_default_avatar(self):
         img_path = modules.get_module_resource('web', 'static/src/img', 'placeholder.png')
         with open(img_path, 'rb') as f:
-            image = f.read()
-        content = base64.b64encode(image)
-        dictheaders = dict(headers) if headers else {}
-        dictheaders['Content-Type'] = 'image/png'
-        if not (width or height):
-            suffix = field.split('_')[-1] if '_' in field else 'large'
-            if suffix in ('small', 'medium', 'large', 'big'):
-                content = getattr(tools, 'image_resize_image_%s' % suffix)(content)
-        return content
+            return base64.b64encode(f.read())
 
     def _check_user_profile_access(self, user_id):
         user_sudo = request.env['res.users'].sudo().browse(user_id)
@@ -80,33 +72,34 @@ class WebsiteProfile(http.Controller):
     @http.route([
         '/profile/avatar/<int:user_id>',
     ], type='http', auth="public", website=True, sitemap=False)
-    def get_user_profile_avatar(self, user_id, field='image_large', width=0, height=0, crop=False, avoid_if_small=False, upper_limit=False, **post):
+    def get_user_profile_avatar(self, user_id, field='image_large', width=0, height=0, crop=False, **post):
         if field not in ('image_small', 'image_medium', 'image_large'):
             return werkzeug.exceptions.Forbidden()
 
         can_sudo = self._check_avatar_access(user_id, **post)
         if can_sudo:
-            status, headers, content = request.env['ir.http'].sudo().binary_content(
+            status, headers, image_base64 = request.env['ir.http'].sudo().binary_content(
                 model='res.users', id=user_id, field=field,
                 default_mimetype='image/png')
         else:
-            status, headers, content = request.env['ir.http'].binary_content(
+            status, headers, image_base64 = request.env['ir.http'].binary_content(
                 model='res.users', id=user_id, field=field,
                 default_mimetype='image/png')
         if status == 301:
-            return request.env['ir.http']._response_by_status(status, headers, content)
+            return request.env['ir.http']._response_by_status(status, headers, image_base64)
         if status == 304:
             return werkzeug.wrappers.Response(status=304)
 
-        if not content:
-            content = self._get_default_avatar(field, headers, width, height)
+        if not image_base64:
+            image_base64 = self._get_default_avatar()
+            if not (width or height):
+                width, height = tools.image_guess_size_from_field_name(field)
 
-        content = tools.limited_image_resize(
-            content, width=width, height=height, crop=crop, upper_limit=upper_limit, avoid_if_small=avoid_if_small)
+        image_base64 = tools.image_process(image_base64, (width, height), crop=crop)
 
-        image_base64 = base64.b64decode(content)
-        headers.append(('Content-Length', len(image_base64)))
-        response = request.make_response(image_base64, headers)
+        content = base64.b64decode(image_base64)
+        headers = http.set_safe_image_headers(headers, content)
+        response = request.make_response(content, headers)
         response.status_code = status
         return response
 
diff --git a/addons/website_sale/tests/test_website_sale_image.py b/addons/website_sale/tests/test_website_sale_image.py
index c873fffef15..3577844e90d 100644
--- a/addons/website_sale/tests/test_website_sale_image.py
+++ b/addons/website_sale/tests/test_website_sale_image.py
@@ -49,19 +49,19 @@ class TestWebsiteSaleImage(odoo.tests.HttpCase):
 
         # first image (blue) for the template
         f = io.BytesIO()
-        Image.new('RGB', (1920, 1080), color_blue).save(f, 'PNG')
+        Image.new('RGB', (1920, 1080), color_blue).save(f, 'JPEG')
         f.seek(0)
         blue_image = base64.b64encode(f.read())
 
         # second image (red) for the variant 1, small image (no zoom)
         f = io.BytesIO()
-        Image.new('RGB', (800, 500), color_red).save(f, 'PNG')
+        Image.new('RGB', (800, 500), color_red).save(f, 'JPEG')
         f.seek(0)
         red_image = base64.b64encode(f.read())
 
         # second image (green) for the variant 2, big image (zoom)
         f = io.BytesIO()
-        Image.new('RGB', (1920, 1080), color_green).save(f, 'PNG')
+        Image.new('RGB', (1920, 1080), color_green).save(f, 'JPEG')
         f.seek(0)
         green_image = base64.b64encode(f.read())
 
@@ -82,9 +82,9 @@ class TestWebsiteSaleImage(odoo.tests.HttpCase):
 
         # Green Variant Extra Image 1
         f = io.BytesIO()
-        Image.new('RGB', (2147, 3251)).save(f, 'JPEG')
+        Image.new('RGB', (2147, 3251)).save(f, 'PNG')
         f.seek(0)
-        image_jpeg = base64.b64encode(f.read())
+        image_png = base64.b64encode(f.read())
 
         # create the template, without creating the variants
         template = self.env['product.template'].with_context(create_product_product=True).create({
@@ -124,7 +124,7 @@ class TestWebsiteSaleImage(odoo.tests.HttpCase):
             'image': green_image,
             'product_tmpl_id': template.id,
             'attribute_value_ids': [(6, 0, attr_values.filtered(lambda l: l.name == name_green).ids)],
-            'product_variant_image_ids': [(0, 0, {'name': 'image 3', 'image': image_jpeg})],
+            'product_variant_image_ids': [(0, 0, {'name': 'image 3', 'image': image_png})],
         })
 
         # now set the red image on the first variant, that works because
@@ -140,65 +140,70 @@ class TestWebsiteSaleImage(odoo.tests.HttpCase):
         self.assertTrue(product_green.can_image_be_zoomed)
         self.assertTrue(product_green.product_variant_image_ids[0].can_image_be_zoomed)
 
+        # jpeg encoding is changing the color a bit
+        jpeg_blue = (65, 105, 227)
+        jpeg_red = (205, 93, 92)
+        jpeg_green = (34, 139, 34)
+
         # Verify original size: keep original
         image = Image.open(io.BytesIO(base64.b64decode(template.image_original)))
         self.assertEqual(image.size, (1920, 1080))
-        self.assertEqual(image.getpixel((image.size[0] / 2, image.size[1] / 2)), (65, 105, 225), "blue")
+        self.assertEqual(image.getpixel((image.size[0] / 2, image.size[1] / 2)), jpeg_blue, "blue")
         image = Image.open(io.BytesIO(base64.b64decode(product_red.image_original)))
         self.assertEqual(image.size, (800, 500))
-        self.assertEqual(image.getpixel((image.size[0] / 2, image.size[1] / 2)), (205, 92, 92), "red")
+        self.assertEqual(image.getpixel((image.size[0] / 2, image.size[1] / 2)), jpeg_red, "red")
         image = Image.open(io.BytesIO(base64.b64decode(product_green.image_original)))
         self.assertEqual(image.size, (1920, 1080))
-        self.assertEqual(image.getpixel((image.size[0] / 2, image.size[1] / 2)), (34, 139, 34), "green")
+        self.assertEqual(image.getpixel((image.size[0] / 2, image.size[1] / 2)), jpeg_green, "green")
 
         # Verify big size: keep aspect ratio
         image = Image.open(io.BytesIO(base64.b64decode(template.image_big)))
         self.assertEqual(image.size, (1024, 576))
-        self.assertEqual(image.getpixel((image.size[0] / 2, image.size[1] / 2)), (65, 105, 225), "blue")
+        self.assertEqual(image.getpixel((image.size[0] / 2, image.size[1] / 2)), jpeg_blue, "blue")
         image = Image.open(io.BytesIO(base64.b64decode(product_red.image_big)))
         self.assertEqual(image.size, (800, 500))
-        self.assertEqual(image.getpixel((image.size[0] / 2, image.size[1] / 2)), (205, 92, 92), "red")
+        self.assertEqual(image.getpixel((image.size[0] / 2, image.size[1] / 2)), jpeg_red, "red")
         image = Image.open(io.BytesIO(base64.b64decode(product_green.image_big)))
         self.assertEqual(image.size, (1024, 576))
-        self.assertEqual(image.getpixel((image.size[0] / 2, image.size[1] / 2)), (34, 139, 34), "green")
+        self.assertEqual(image.getpixel((image.size[0] / 2, image.size[1] / 2)), jpeg_green, "green")
 
         # Verify image == image_big
         self.assertEqual(template.image_big, template.image)
         self.assertEqual(product_red.image_big, product_red.image)
         self.assertEqual(product_green.image_big, product_green.image)
 
-        # Verify large size
+        # Verify large size: keep aspect ratio
         image = Image.open(io.BytesIO(base64.b64decode(template.image_large)))
-        self.assertEqual(image.size, (256, 256))
-        self.assertEqual(image.getpixel((image.size[0] / 2, image.size[1] / 2)), (65, 105, 225), "blue")
+        self.assertEqual(image.size, (256, 144))
+        self.assertEqual(image.getpixel((image.size[0] / 2, image.size[1] / 2)), jpeg_blue, "blue")
         image = Image.open(io.BytesIO(base64.b64decode(product_red.image_large)))
-        self.assertEqual(image.size, (256, 256))
-        self.assertEqual(image.getpixel((image.size[0] / 2, image.size[1] / 2)), (205, 92, 92), "red")
+        self.assertEqual(image.size, (256, 160))
+        self.assertEqual(image.getpixel((image.size[0] / 2, image.size[1] / 2)), jpeg_red, "red")
         image = Image.open(io.BytesIO(base64.b64decode(product_green.image_large)))
-        self.assertEqual(image.size, (256, 256))
-        self.assertEqual(image.getpixel((image.size[0] / 2, image.size[1] / 2)), (34, 139, 34), "green")
+        self.assertEqual(image.size, (256, 144))
+        self.assertEqual(image.getpixel((image.size[0] / 2, image.size[1] / 2)), jpeg_green, "green")
 
-        # Verify medium size
+        # Verify medium size: keep aspect ratio
         image = Image.open(io.BytesIO(base64.b64decode(template.image_medium)))
-        self.assertEqual(image.size, (128, 128))
-        self.assertEqual(image.getpixel((image.size[0] / 2, image.size[1] / 2)), (65, 105, 225), "blue")
+        self.assertEqual(image.size, (128, 72))
+        self.assertEqual(image.getpixel((image.size[0] / 2, image.size[1] / 2)), jpeg_blue, "blue")
         image = Image.open(io.BytesIO(base64.b64decode(product_red.image_medium)))
-        self.assertEqual(image.size, (128, 128))
-        self.assertEqual(image.getpixel((image.size[0] / 2, image.size[1] / 2)), (205, 92, 92), "red")
+        self.assertEqual(image.size, (128, 80))
+        self.assertEqual(image.getpixel((image.size[0] / 2, image.size[1] / 2)), jpeg_red, "red")
         image = Image.open(io.BytesIO(base64.b64decode(product_green.image_medium)))
-        self.assertEqual(image.size, (128, 128))
-        self.assertEqual(image.getpixel((image.size[0] / 2, image.size[1] / 2)), (34, 139, 34), "green")
+        self.assertEqual(image.size, (128, 72))
+        self.assertEqual(image.getpixel((image.size[0] / 2, image.size[1] / 2)), jpeg_green, "green")
 
-        # Verify small size
+        # Verify small size: keep aspect ratio
         image = Image.open(io.BytesIO(base64.b64decode(template.image_small)))
-        self.assertEqual(image.size, (64, 64))
-        self.assertEqual(image.getpixel((image.size[0] / 2, image.size[1] / 2)), (65, 105, 225), "blue")
+        self.assertEqual(image.size, (64, 36))
+        self.assertEqual(image.getpixel((image.size[0] / 2, image.size[1] / 2)), jpeg_blue, "blue")
         image = Image.open(io.BytesIO(base64.b64decode(product_red.image_small)))
-        self.assertEqual(image.size, (64, 64))
-        self.assertEqual(image.getpixel((image.size[0] / 2, image.size[1] / 2)), (205, 92, 92), "red")
+        self.assertEqual(image.size, (64, 40))
+        self.assertEqual(image.getpixel((image.size[0] / 2, image.size[1] / 2)), jpeg_red, "red")
         image = Image.open(io.BytesIO(base64.b64decode(product_green.image_small)))
-        self.assertEqual(image.size, (64, 64))
-        self.assertEqual(image.getpixel((image.size[0] / 2, image.size[1] / 2)), (34, 139, 34), "green")
+        self.assertEqual(image.size, (64, 36))
+        self.assertEqual(image.getpixel((image.size[0] / 2, image.size[1] / 2)), jpeg_green, "green")
 
         # self.env.cr.commit()  # uncomment to save the product to test in browser
 
@@ -223,7 +228,7 @@ class TestWebsiteSaleImage(odoo.tests.HttpCase):
         # write on the raw field, otherwise it will write on the fallback here
         product_green.image_raw_original = False
         images = product_green._get_images()
-        self.assertEqual(images[0].image_original, image_jpeg)
+        self.assertEqual(images[0].image_original, image_png)
         self.assertEqual(images[1].image_original, red_image)
         self.assertEqual(images[2].image_original, image_gif)
         self.assertEqual(images[3].image_original, image_svg)
diff --git a/addons/website_sale/views/templates.xml b/addons/website_sale/views/templates.xml
index 302e2654f7b..beaa97634d9 100644
--- a/addons/website_sale/views/templates.xml
+++ b/addons/website_sale/views/templates.xml
@@ -933,7 +933,7 @@
         <t t-if="website_sale_order and website_sale_order.website_order_line">
             <t t-foreach="website_sale_order.website_order_line" t-as="line">
                 <div class="row mb8 cart_line">
-                    <div class="col-3">
+                    <div class="col-3 text-center">
                         <span t-field="line.product_id.image_small" t-options="{'widget': 'image', 'class': 'rounded'}" />
                     </div>
                     <div class="col-9">
@@ -1414,7 +1414,7 @@
                         <tbody>
                             <tr t-foreach="website_sale_order.website_order_line" t-as="line">
                                 <td class='' colspan="2" t-if="not line.product_id.product_tmpl_id"></td>
-                                <td class='td-img' t-if="line.product_id.product_tmpl_id">
+                                <td class='td-img text-center' t-if="line.product_id.product_tmpl_id">
                                     <span t-field="line.product_id.image_small" t-options="{'widget': 'image', 'class': 'rounded'}" />
                                 </td>
                                 <td class='td-product_name' t-if="line.product_id.product_tmpl_id">
diff --git a/addons/website_slides/controllers/main.py b/addons/website_slides/controllers/main.py
index edb00cc0e7b..4da2d91f6a7 100644
--- a/addons/website_slides/controllers/main.py
+++ b/addons/website_slides/controllers/main.py
@@ -506,7 +506,7 @@ class WebsiteSlides(WebsiteProfile):
         return response
 
     @http.route('/slides/slide/<int:slide_id>/get_image', type='http', auth="public", website=True, sitemap=False)
-    def slide_get_image(self, slide_id, field='image_medium', width=0, height=0, crop=False, avoid_if_small=False, upper_limit=False):
+    def slide_get_image(self, slide_id, field='image_medium', width=0, height=0, crop=False):
         # Protect infographics by limiting access to 256px (large) images
         if field not in ('image_small', 'image_medium', 'image_large'):
             return werkzeug.exceptions.Forbidden()
@@ -515,23 +515,24 @@ class WebsiteSlides(WebsiteProfile):
         if not slide:
             raise werkzeug.exceptions.NotFound()
 
-        status, headers, content = request.env['ir.http'].sudo().binary_content(
+        status, headers, image_base64 = request.env['ir.http'].sudo().binary_content(
             model='slide.slide', id=slide.id, field=field,
             default_mimetype='image/png')
         if status == 301:
-            return request.env['ir.http']._response_by_status(status, headers, content)
+            return request.env['ir.http']._response_by_status(status, headers, image_base64)
         if status == 304:
             return werkzeug.wrappers.Response(status=304)
 
-        if not content:
-            content = self._get_default_avatar(field, headers, width, height)
+        if not image_base64:
+            image_base64 = self._get_default_avatar()
+            if not (width or height):
+                width, height = tools.image_guess_size_from_field_name(field)
 
-        content = tools.limited_image_resize(
-            content, width=width, height=height, crop=crop, upper_limit=upper_limit, avoid_if_small=avoid_if_small)
+        image_base64 = tools.image_process(image_base64, (width, height), crop=crop)
 
-        image_base64 = base64.b64decode(content)
-        headers.append(('Content-Length', len(image_base64)))
-        response = request.make_response(image_base64, headers)
+        content = base64.b64decode(image_base64)
+        headers = http.set_safe_image_headers(headers, content)
+        response = request.make_response(content, headers)
         response.status_code = status
         return response
 
diff --git a/odoo/addons/base/models/image_mixin.py b/odoo/addons/base/models/image_mixin.py
index 9ef70b15bbd..1b186c993b4 100644
--- a/odoo/addons/base/models/image_mixin.py
+++ b/odoo/addons/base/models/image_mixin.py
@@ -28,7 +28,7 @@ class ImageMixin(models.AbstractModel):
         for record in self:
             images = tools.image_get_resized_images(record.image_original, big_name=False)
             record.image_big = tools.image_get_resized_images(record.image_original,
-                large_name=False, medium_name=False, small_name=False, preserve_aspect_ratio=True)['image']
+                large_name=False, medium_name=False, small_name=False)['image']
             record.image_large = images['image_large']
             record.image_medium = images['image_medium']
             record.image_small = images['image_small']
diff --git a/odoo/addons/base/models/res_company.py b/odoo/addons/base/models/res_company.py
index 8925f6044b5..db64a8265a3 100644
--- a/odoo/addons/base/models/res_company.py
+++ b/odoo/addons/base/models/res_company.py
@@ -120,7 +120,7 @@ class Company(models.Model):
     @api.depends('partner_id', 'partner_id.image')
     def _compute_logo_web(self):
         for company in self:
-            company.logo_web = tools.image_resize_image(company.partner_id.image, (180, None))
+            company.logo_web = tools.image_process(company.partner_id.image, (180, None))
 
     @api.onchange('state_id')
     def _onchange_state(self):
diff --git a/odoo/addons/base/models/res_partner.py b/odoo/addons/base/models/res_partner.py
index 49979b9e2c5..dd7635d5312 100644
--- a/odoo/addons/base/models/res_partner.py
+++ b/odoo/addons/base/models/res_partner.py
@@ -328,29 +328,29 @@ class Partner(models.Model):
         if getattr(threading.currentThread(), 'testing', False) or self._context.get('install_mode'):
             return False
 
-        colorize, img_path, image = False, False, False
+        colorize, img_path, image_base64 = False, False, False
 
         if partner_type in ['other'] and parent_id:
             parent_image = self.browse(parent_id).image
-            image = parent_image and base64.b64decode(parent_image) or None
+            image_base64 = parent_image or None
 
-        if not image and partner_type == 'invoice':
+        if not image_base64 and partner_type == 'invoice':
             img_path = get_module_resource('base', 'static/img', 'money.png')
-        elif not image and partner_type == 'delivery':
+        elif not image_base64 and partner_type == 'delivery':
             img_path = get_module_resource('base', 'static/img', 'truck.png')
-        elif not image and is_company:
+        elif not image_base64 and is_company:
             img_path = get_module_resource('base', 'static/img', 'company_image.png')
-        elif not image:
+        elif not image_base64:
             img_path = get_module_resource('base', 'static/img', 'avatar.png')
             colorize = True
 
         if img_path:
             with open(img_path, 'rb') as f:
-                image = f.read()
-        if image and colorize:
-            image = tools.image_colorize(image)
+                image_base64 = base64.b64encode(f.read())
+        if image_base64 and colorize:
+            image_base64 = tools.image_process(image_base64, colorize=True)
 
-        return tools.image_resize_image_big(base64.b64encode(image))
+        return tools.image_process(image_base64, size=tools.IMAGE_BIG_SIZE)
 
     @api.model
     def _fields_view_get(self, view_id=None, view_type='form', toolbar=False, submenu=False):
@@ -572,7 +572,7 @@ class Partner(models.Model):
                             ("The selected company is not compatible with the companies of the related user(s)"))
         # no padding on the big image, because it's used as website logo
         tools.image_resize_images(vals, return_big=False)
-        tools.image_resize_images(vals, return_medium=False, return_small=False, preserve_aspect_ratio=True)
+        tools.image_resize_images(vals, return_medium=False, return_small=False)
 
         result = True
         # To write in SUPERUSER on field is_company and avoid access rights problems.
@@ -601,7 +601,7 @@ class Partner(models.Model):
                 vals['image'] = self._get_default_image(vals.get('type'), vals.get('is_company'), vals.get('parent_id'))
             # no padding on the big image, because it's used as website logo
             tools.image_resize_images(vals, return_big=False)
-            tools.image_resize_images(vals, return_medium=False, return_small=False, preserve_aspect_ratio=True)
+            tools.image_resize_images(vals, return_medium=False, return_small=False)
         partners = super(Partner, self).create(vals_list)
 
         if self.env.context.get('_partners_skip_fields_sync'):
diff --git a/odoo/addons/base/static/img/avatar_grey.png b/odoo/addons/base/static/img/avatar_grey.png
new file mode 100644
index 00000000000..db05e744f7c
Binary files /dev/null and b/odoo/addons/base/static/img/avatar_grey.png differ
diff --git a/odoo/addons/base/tests/__init__.py b/odoo/addons/base/tests/__init__.py
index 281aec864c8..bba0d5b6742 100644
--- a/odoo/addons/base/tests/__init__.py
+++ b/odoo/addons/base/tests/__init__.py
@@ -7,6 +7,7 @@ from . import test_db_cursor
 from . import test_expression
 from . import test_float
 from . import test_func
+from . import test_image
 from . import test_ir_actions
 from . import test_ir_attachment
 from . import test_ir_http
diff --git a/odoo/addons/base/tests/test_image.py b/odoo/addons/base/tests/test_image.py
new file mode 100644
index 00000000000..b9b77dcb2a0
--- /dev/null
+++ b/odoo/addons/base/tests/test_image.py
@@ -0,0 +1,235 @@
+# -*- coding: utf-8 -*-
+# Part of Odoo. See LICENSE file for full copyright and licensing details.
+
+import base64
+import binascii
+from PIL import Image, ImageDraw, PngImagePlugin
+
+from odoo import tools
+from odoo.tests.common import TransactionCase
+
+
+class TestImage(TransactionCase):
+    """Tests for the different image tools helpers.
+
+    The following helpers are not tested here because they are wrappers for the
+    other methods or they are tested elsewhere (eg. on TestWebsiteSaleImage):
+    - image_get_resized_images
+    - image_resize_images
+    - is_image_size_above
+    """
+    def setUp(self):
+        super(TestImage, self).setUp()
+        self.bg_color = (135, 90, 123)
+        self.fill_color = (0, 160, 157)
+
+        self.base64_1x1_png = b'iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAIAAACQd1PeAAAADElEQVR4nGNgYGAAAAAEAAH2FzhVAAAAAElFTkSuQmCC'
+        self.base64_svg = base64.b64encode(b'<svg></svg>')
+        self.base64_1920x1080_jpeg = tools.image_to_base64(Image.new('RGB', (1920, 1080)), 'JPEG')
+
+        # Draw a red square in the middle of the image, this will be used to
+        # verify crop is working. The border is going to be `self.bg_color` and
+        # the middle is going to be `self.fill_color`.
+
+        # horizontal image (border is left/right)
+        image = Image.new('RGB', (1920, 1080), color=self.bg_color)
+        offset = (image.size[0] - image.size[1]) / 2
+        draw = ImageDraw.Draw(image)
+        draw.rectangle(xy=[
+            (offset, 0),
+            (image.size[0] - offset, image.size[1])
+        ], fill=self.fill_color)
+        self.base64_1920x1080_png = tools.image_to_base64(image, 'PNG')
+
+        # vertical image (border is top/bottom)
+        image = Image.new('RGB', (1080, 1920), color=self.bg_color)
+        offset = (image.size[1] - image.size[0]) / 2
+        draw = ImageDraw.Draw(image)
+        draw.rectangle(xy=[
+            (0, offset),
+            (image.size[0], image.size[1] - offset)
+        ], fill=self.fill_color)
+        self.base64_1080x1920_png = tools.image_to_base64(image, 'PNG')
+
+    def test_00_base64_to_image(self):
+        """Test that base64 is correctly opened as a PIL image."""
+        image = tools.base64_to_image(self.base64_1x1_png)
+        self.assertEqual(type(image), PngImagePlugin.PngImageFile, "base64 as bytes, correct format")
+        self.assertEqual(image.size, (1, 1), "base64 as bytes, correct size")
+
+        image = tools.base64_to_image(self.base64_1x1_png.decode('ASCII'))
+        self.assertEqual(type(image), PngImagePlugin.PngImageFile, "base64 as string, correct format")
+        self.assertEqual(image.size, (1, 1), "base64 as string, correct size")
+
+        with self.assertRaises(binascii.Error, msg="wrong base64: binascii.Error: Incorrect padding"):
+            image = tools.base64_to_image(b'oazdazpodazdpok')
+
+        with self.assertRaises(OSError, msg="wrong base64: OSError: cannot identify image file"):
+            image = tools.base64_to_image(b'oazdazpodazdpokd')
+
+    def test_01_image_to_base64(self):
+        """Test that a PIL image is correctly saved as base64."""
+        image = Image.new('RGB', (1, 1))
+        image_base64 = tools.image_to_base64(image, 'PNG')
+        self.assertEqual(image_base64, self.base64_1x1_png)
+
+    def test_10_image_process_base64_source(self):
+        """Test the base64_source parameter of image_process."""
+        self.assertFalse(tools.image_process(False), "return False if base64_source is falsy")
+        self.assertEqual(tools.image_process(self.base64_svg), self.base64_svg, "return base64_source if format is SVG")
+
+        with self.assertRaises(binascii.Error, msg="wrong base64: binascii.Error: Incorrect padding"):
+            tools.image_process(b'oazdazpodazdpok')
+
+        with self.assertRaises(OSError, msg="wrong base64: OSError: cannot identify image file"):
+            tools.image_process(b'oazdazpodazdpokd')
+
+        image = tools.base64_to_image(tools.image_process(self.base64_1920x1080_jpeg))
+        self.assertEqual(image.size, (1920, 1080), "OK return the image")
+
+    def test_11_image_process_size(self):
+        """Test the size parameter of image_process."""
+
+        # Format of `tests`: (original base64 image, size parameter, expected result, text)
+        tests = [
+            (self.base64_1920x1080_jpeg, (192, 108), (192, 108), "resize to given size"),
+            (self.base64_1920x1080_jpeg, (1920, 1080), (1920, 1080), "same size, no change"),
+            (self.base64_1920x1080_jpeg, (192, None), (192, 108), "set height from ratio"),
+            (self.base64_1920x1080_jpeg, (0, 108), (192, 108), "set width from ratio"),
+            (self.base64_1920x1080_jpeg, (192, 200), (192, 108), "adapt to width"),
+            (self.base64_1920x1080_jpeg, (400, 108), (192, 108), "adapt to height"),
+            (self.base64_1920x1080_jpeg, (3000, 2000), (1920, 1080), "don't resize above original, both set"),
+            (self.base64_1920x1080_jpeg, (3000, False), (1920, 1080), "don't resize above original, width set"),
+            (self.base64_1920x1080_jpeg, (None, 2000), (1920, 1080), "don't resize above original, height set"),
+            (self.base64_1080x1920_png, (3000, 192), (108, 192), "vertical image, resize if below"),
+        ]
+
+        count = 0
+        for test in tests:
+            image = tools.base64_to_image(tools.image_process(test[0], size=test[1]))
+            self.assertEqual(image.size, test[2], test[3])
+            count = count + 1
+        self.assertEqual(count, 10, "ensure the loop is ran")
+
+    def test_12_image_process_verify_resolution(self):
+        """Test the verify_resolution parameter of image_process."""
+        res = tools.image_process(self.base64_1920x1080_jpeg, verify_resolution=True)
+        self.assertNotEqual(res, False, "size ok")
+        base64_image_excessive = tools.image_to_base64(Image.new('RGB', (45001, 1000)), 'PNG')
+        with self.assertRaises(ValueError, msg="size excessive"):
+            tools.image_process(base64_image_excessive, verify_resolution=True)
+
+    def test_13_image_process_quality(self):
+        """Test the quality parameter of image_process."""
+
+        # CASE: PNG RGBA doesn't apply quality, just optimize
+        image = tools.image_to_base64(Image.new('RGBA', (1080, 1920)), 'PNG')
+        res = tools.image_process(image)
+        self.assertLessEqual(len(res), len(image))
+
+        # CASE: PNG RGB doesn't apply quality, just optimize
+        image = tools.image_to_base64(Image.new('P', (1080, 1920)), 'PNG')
+        res = tools.image_process(image)
+        self.assertLessEqual(len(res), len(image))
+
+        # CASE: JPEG optimize + reduced quality
+        res = tools.image_process(self.base64_1920x1080_jpeg)
+        self.assertLessEqual(len(res), len(self.base64_1920x1080_jpeg))
+
+        # CASE: GIF doesn't apply quality, just optimize
+        image = tools.image_to_base64(Image.new('RGB', (1080, 1920)), 'GIF')
+        res = tools.image_process(image)
+        self.assertLessEqual(len(res), len(image))
+
+    def test_14_image_process_crop(self):
+        """Test the crop parameter of image_process."""
+
+        # Optimized PNG use palette, getpixel below will return palette value.
+        fill = 0
+        bg = 1
+
+        # Format of `tests`: (original base64 image, size parameter, crop parameter, res size, res color (top, bottom, left, right), text)
+        tests = [
+            (self.base64_1920x1080_png, None, None, (1920, 1080), (fill, fill, bg, bg), "horizontal, verify initial"),
+            (self.base64_1920x1080_png, (2000, 2000), 'center', (1080, 1080), (fill, fill, fill, fill), "horizontal, crop biggest possible"),
+            (self.base64_1920x1080_png, (2000, 4000), 'center', (540, 1080), (fill, fill, fill, fill), "horizontal, size vertical, limit height"),
+            (self.base64_1920x1080_png, (4000, 2000), 'center', (1920, 960), (fill, fill, bg, bg), "horizontal, size horizontal, limit width"),
+            (self.base64_1920x1080_png, (512, 512), 'center', (512, 512), (fill, fill, fill, fill), "horizontal, type center"),
+            (self.base64_1920x1080_png, (512, 512), 'top', (512, 512), (fill, fill, fill, fill), "horizontal, type top"),
+            (self.base64_1920x1080_png, (512, 512), 'bottom', (512, 512), (fill, fill, fill, fill), "horizontal, type bottom"),
+            (self.base64_1920x1080_png, (512, 512), 'wrong', (512, 512), (fill, fill, fill, fill), "horizontal, wrong crop value, use center"),
+            (self.base64_1920x1080_png, (192, 0), None, (192, 108), (fill, fill, bg, bg), "horizontal, not cropped, just do resize"),
+
+            (self.base64_1080x1920_png, None, None, (1080, 1920), (bg, bg, fill, fill), "vertical, verify initial"),
+            (self.base64_1080x1920_png, (2000, 2000), 'center', (1080, 1080), (fill, fill, fill, fill), "vertical, crop biggest possible"),
+            (self.base64_1080x1920_png, (2000, 4000), 'center', (960, 1920), (bg, bg, fill, fill), "vertical, size vertical, limit height"),
+            (self.base64_1080x1920_png, (4000, 2000), 'center', (1080, 540), (fill, fill, fill, fill), "vertical, size horizontal, limit width"),
+            (self.base64_1080x1920_png, (512, 512), 'center', (512, 512), (fill, fill, fill, fill), "vertical, type center"),
+            (self.base64_1080x1920_png, (512, 512), 'top', (512, 512), (bg, fill, fill, fill), "vertical, type top"),
+            (self.base64_1080x1920_png, (512, 512), 'bottom', (512, 512), (fill, bg, fill, fill), "vertical, type bottom"),
+            (self.base64_1080x1920_png, (512, 512), 'wrong', (512, 512), (fill, fill, fill, fill), "vertical, wrong crop value, use center"),
+            (self.base64_1080x1920_png, (108, 0), None, (108, 192), (bg, bg, fill, fill), "vertical, not cropped, just do resize"),
+        ]
+
+        count = 0
+        for test in tests:
+            count = count + 1
+            # process the image
+            image = tools.base64_to_image(tools.image_process(test[0], size=test[1], crop=test[2]))
+            # verify size
+            self.assertEqual(image.size, test[3], "%s - correct size" % test[5])
+
+            half_width, half_height = image.size[0] / 2, image.size[1] / 2
+            top, bottom, left, right = 0, image.size[1] - 1, 0, image.size[0] - 1
+            # verify top
+            px = (half_width, top)
+            self.assertEqual(image.getpixel(px), test[4][0], "%s - color top (%s, %s)" % (test[5], px[0], px[1]))
+            # verify bottom
+            px = (half_width, bottom)
+            self.assertEqual(image.getpixel(px), test[4][1], "%s - color bottom (%s, %s)" % (test[5], px[0], px[1]))
+            # verify left
+            px = (left, half_height)
+            self.assertEqual(image.getpixel(px), test[4][2], "%s - color left (%s, %s)" % (test[5], px[0], px[1]))
+            # verify right
+            px = (right, half_height)
+            self.assertEqual(image.getpixel(px), test[4][3], "%s - color right (%s, %s)" % (test[5], px[0], px[1]))
+
+        self.assertEqual(count, 2 * 9, "ensure the loop is ran")
+
+    def test_15_image_process_colorize(self):
+        """Test the colorize parameter of image_process."""
+
+        # verify initial condition
+        image_rgba = Image.new('RGBA', (1, 1))
+        self.assertEqual(image_rgba.mode, 'RGBA')
+        self.assertEqual(image_rgba.getpixel((0, 0)), (0, 0, 0, 0))
+        base64_rgba = tools.image_to_base64(image_rgba, 'PNG')
+
+        # CASE: color random, color has changed
+        image = tools.base64_to_image(tools.image_process(base64_rgba, colorize=True))
+        self.assertEqual(image.mode, 'P')
+        self.assertNotEqual(image.getpixel((0, 0)), (0, 0, 0))
+
+    def test_16_image_process_format(self):
+        """Test the format parameter of image_process."""
+
+        image = tools.base64_to_image(tools.image_process(self.base64_1920x1080_jpeg, output_format='PNG'))
+        self.assertEqual(image.format, 'PNG', "change format to PNG")
+
+        image = tools.base64_to_image(tools.image_process(self.base64_1x1_png, output_format='JpEg'))
+        self.assertEqual(image.format, 'JPEG', "change format to JPEG (case insensitive)")
+
+        image = tools.base64_to_image(tools.image_process(self.base64_1920x1080_jpeg, output_format='BMP'))
+        self.assertEqual(image.format, 'PNG', "change format to BMP converted to PNG")
+
+        self.base64_image_1080_1920_rgba = tools.image_to_base64(Image.new('RGBA', (108, 192)), 'PNG')
+        image = tools.base64_to_image(tools.image_process(self.base64_image_1080_1920_rgba, output_format='jpeg'))
+        self.assertEqual(image.format, 'JPEG', "change format PNG with RGBA to JPEG")
+
+        self.base64_image_1080_1920_tiff = tools.image_to_base64(Image.new('RGB', (108, 192)), 'TIFF')
+        image = tools.base64_to_image(tools.image_process(self.base64_image_1080_1920_tiff))
+        self.assertEqual(image.format, 'JPEG', "unsupported format to JPEG")
+
+    def test_20_image_data_uri(self):
+        """Test that image_data_uri is working as expected."""
+        self.assertEqual(tools.image_data_uri(self.base64_1x1_png), 'data:image/png;base64,' + self.base64_1x1_png.decode('ascii'))
diff --git a/odoo/addons/base/tests/test_ir_http.py b/odoo/addons/base/tests/test_ir_http.py
index b332a433302..9f034c9d896 100644
--- a/odoo/addons/base/tests/test_ir_http.py
+++ b/odoo/addons/base/tests/test_ir_http.py
@@ -64,7 +64,7 @@ class test_ir_http_mimetype(common.TransactionCase):
             'type': 'binary',
         })
 
-        resized = odoo.tools.image_resize_image_small(prop.value_binary)
+        resized = odoo.tools.image_process(prop.value_binary, odoo.tools.IMAGE_SMALL_SIZE)
         # Simul computed field which resize and that is not attachement=True (E.G. on product)
         prop.write({'value_binary': resized})
         status, headers, content = self.env['ir.http'].binary_content(
diff --git a/odoo/addons/base/views/res_partner_views.xml b/odoo/addons/base/views/res_partner_views.xml
index d11fa33b479..ffb6395a949 100644
--- a/odoo/addons/base/views/res_partner_views.xml
+++ b/odoo/addons/base/views/res_partner_views.xml
@@ -502,37 +502,25 @@
                     <field name="mobile"/>
                     <field name="state_id"/>
                     <field name="category_id"/>
-                    <field name="image_small"/>
+                    <field name="image_medium"/>
                     <field name="type"/>
                     <templates>
                         <t t-name="kanban-box">
                             <div class="oe_kanban_global_click o_kanban_record_has_image_fill o_res_partner_kanban">
-                                <t t-if="record.image_small.raw_value &amp;&amp; !record.is_company.raw_value">
-                                    <div class="o_kanban_image_fill_left d-none d-md-block" t-attf-style="background-image:url('#{kanban_image('res.partner', 'image_medium', record.id.raw_value)}')">
+                                <t t-if="!record.is_company.raw_value">
+                                    <t t-if="record.type.raw_value === 'delivery'" t-set="placeholder" t-value="'/base/static/img/truck.png'"/>
+                                    <t t-elif="record.type.raw_value === 'invoice'" t-set="placeholder" t-value="'/base/static/img/money.png'"/>
+                                    <t t-else="" t-set="placeholder" t-value="'/base/static/img/avatar_grey.png'"/>
+                                    <div class="o_kanban_image_fill_left d-none d-md-block" t-attf-style="background-image:url('#{kanban_image('res.partner', 'image_medium', record.id.raw_value,  placeholder)}')">
                                         <img class="o_kanban_image_inner_pic" t-if="record.parent_id.raw_value" t-att-alt="record.parent_id.value" t-att-src="kanban_image('res.partner', 'image_small', record.parent_id.raw_value)"/>
                                     </div>
-                                    <div class="o_kanban_image rounded-circle d-md-none" t-attf-style="background-image:url('#{kanban_image('res.partner', 'image_small', record.id.raw_value)}')">
+                                    <div class="o_kanban_image rounded-circle d-md-none" t-attf-style="background-image:url('#{kanban_image('res.partner', 'image_medium', record.id.raw_value,  placeholder)}')">
                                         <img class="o_kanban_image_inner_pic" t-if="record.parent_id.raw_value" t-att-alt="record.parent_id.value" t-att-src="kanban_image('res.partner', 'image_small', record.parent_id.raw_value)"/>
                                     </div>
                                 </t>
-                                <t t-elif="record.is_company.raw_value">
-                                    <img class="o_kanban_image" t-att-src="kanban_image('res.partner', 'image_small', record.id.raw_value)" t-att-alt="record.display_name"/>
-                                </t>
                                 <t t-else="">
-                                    <t t-if="record.type.raw_value === 'delivery'">
-                                        <img alt="Delivery" t-att-src='_s + "/base/static/img/truck.png"' class="o_kanban_image oe_kanban_avatar_smallbox"/>
-                                    </t>
-                                    <t t-if="record.type.raw_value === 'invoice'">
-                                        <img alt="Invoice" t-att-src='_s + "/base/static/img/money.png"' class="o_kanban_image oe_kanban_avatar_smallbox"/>
-                                    </t>
-                                    <t t-if="record.type.raw_value != 'invoice' &amp;&amp; record.type.raw_value != 'delivery'">
-                                        <t t-if="record.is_company.raw_value === true">
-                                            <img class="o_kanban_image" alt="Logo" t-att-src='_s + "/base/static/img/company_image.png"'/>
-                                        </t>
-                                        <t t-if="record.is_company.raw_value === false">
-                                            <img class="o_kanban_image" alt="Avatar" t-att-src='_s + "/base/static/img/avatar.png"'/>
-                                        </t>
-                                    </t>
+                                    <t t-set="placeholder" t-value="'/base/static/img/company_image.png'"/>
+                                    <div class="o_kanban_image_fill_left o_kanban_image_full" t-attf-style="background-image: url(#{kanban_image('res.partner', 'image_medium', record.id.raw_value, placeholder)})" role="img"/>
                                 </t>
                                 <div class="oe_kanban_details">
                                     <strong class="o_kanban_record_title oe_partner_heading"><field name="display_name"/></strong>
diff --git a/odoo/http.py b/odoo/http.py
index ddb18eb1048..750c0b3c0d5 100644
--- a/odoo/http.py
+++ b/odoo/http.py
@@ -50,6 +50,7 @@ from .service.server import memory_info
 from .service import security, model as service_model
 from .tools.func import lazy_property
 from .tools import ustr, consteq, frozendict, pycompat, unique, date_utils
+from .tools.mimetypes import guess_mimetype
 
 from .modules.module import module_manifest
 
@@ -1617,5 +1618,39 @@ def content_disposition(filename):
 
     return "attachment; filename*=UTF-8''%s" % escaped
 
+
+def set_safe_image_headers(headers, content):
+    """Return new headers based on `headers` but with `Content-Length` and
+    `Content-Type` set appropriately depending on the given `content` only if it
+    is safe to do."""
+    content_type = guess_mimetype(content)
+    safe_types = ['image/jpeg', 'image/png', 'image/gif', 'image/x-icon']
+    if content_type in safe_types:
+        headers = set_header_field(headers, 'Content-Type', content_type)
+    set_header_field(headers, 'Content-Length', len(content))
+    return headers
+
+
+def set_header_field(headers, name, value):
+    """ Return new headers based on `headers` but with `value` set for the
+    header field `name`.
+
+    :param headers: the existing headers
+    :type headers: list of tuples (name, value)
+
+    :param name: the header field name
+    :type name: string
+
+    :param value: the value to set for the `name` header
+    :type value: string
+
+    :return: the updated headers
+    :rtype: list of tuples (name, value)
+    """
+    dictheaders = dict(headers)
+    dictheaders[name] = value
+    return list(dictheaders.items())
+
+
 #  main wsgi handler
 root = Root()
diff --git a/odoo/tools/image.py b/odoo/tools/image.py
index 137f84d1ab4..7bb77191e87 100644
--- a/odoo/tools/image.py
+++ b/odoo/tools/image.py
@@ -1,13 +1,17 @@
 # -*- coding: utf-8 -*-
 # Part of Odoo. See LICENSE file for full copyright and licensing details.
 import base64
-import codecs
 import io
 
 from PIL import Image
-from PIL import ImageEnhance
+# We can preload Ico too because it is considered safe
+from PIL import IcoImagePlugin
+
 from random import randrange
 
+from odoo.tools.translate import _
+
+
 # Preload PIL with the minimal subset of image formats we need
 Image.preinit()
 Image._initialized = 2
@@ -26,285 +30,297 @@ IMAGE_LARGE_SIZE = (256, 256)
 IMAGE_MEDIUM_SIZE = (128, 128)
 IMAGE_SMALL_SIZE = (64, 64)
 
+# Arbitraty limit to fit most resolutions, including Nokia Lumia 1020 photo,
+# 8K with a ratio up to 16:10, and almost all variants of 4320p
+IMAGE_MAX_RESOLUTION = 45e6
 
-# ----------------------------------------
-# Image resizing
-# ----------------------------------------
 
-def image_resize_image(base64_source, size=IMAGE_BIG_SIZE, encoding='base64', filetype=None, avoid_if_small=False, preserve_aspect_ratio=False, upper_limit=False):
-    """ Function to resize an image. The image will be resized to the given
-        size, while keeping the aspect ratios, and holes in the image will be
-        filled with transparent background. The image will not be stretched if
-        smaller than the expected size.
-        Steps of the resizing:
-        - Compute width and height if not specified.
-        - if avoid_if_small: if both image sizes are smaller than the requested
-          sizes, the original image is returned. This is used to avoid adding
-          transparent content around images that we do not want to alter but
-          just resize if too big. This is used for example when storing images
-          in the 'image' field: we keep the original image, resized to a maximal
-          size, without adding transparent content around it if smaller.
-        - create a thumbnail of the source image through using the thumbnail
-          function. Aspect ratios are preserved when using it. Note that if the
-          source image is smaller than the expected size, it will not be
-          extended, but filled to match the size.
-        - create a transparent background that will hold the final image.
-        - paste the thumbnail on the transparent background and center it.
+class ImageProcess():
 
-        :param base64_source: base64-encoded version of the source
-            image; if False, returns False
-        :param size: 2-tuple(width, height). A None value for any of width or
-            height mean an automatically computed value based respectively
-            on height or width of the source image.
-        :param encoding: the output encoding
-        :param filetype: the output filetype, by default the source image's
-        :type filetype: str, any PIL image format (supported for creation)
-        :param avoid_if_small: do not resize if image height and width
-            are smaller than the expected size.
-    """
-    if not base64_source:
-        return False
-    # Return unmodified content if no resize or we etect first 6 bits of '<'
-    # (0x3C) for SVG documents - This will bypass XML files as well, but it's
-    # harmless for these purposes
-    if size == (None, None) or base64_source[:1] == b'P':
-        return base64_source
-    image_stream = io.BytesIO(codecs.decode(base64_source, encoding))
-    image = Image.open(image_stream)
-    # store filetype here, as Image.new below will lose image.format
-    filetype = (filetype or image.format).upper()
+    def __init__(self, base64_source, verify_resolution=True):
+        """Initialize the `base64_source` image for processing.
 
-    filetype = {
-        'BMP': 'PNG',
-    }.get(filetype, filetype)
+        :param base64_source: the original image base64 encoded
+            No processing will be done if the `base64_source` is falsy or if
+            the image is SVG.
+        :type base64_source: string or bytes
 
-    asked_width, asked_height = size
-    if upper_limit:
-        if asked_width:
-            asked_width = min(asked_width, image.size[0])
-        if asked_height:
-            asked_height = min(asked_height, image.size[1])
+        :param verify_resolution: if True, make sure the original image size is not
+            excessive before starting to process it. The max allowed resolution is
+            defined by `IMAGE_MAX_RESOLUTION`.
+        :type verify_resolution: bool
 
-        if image.size[0] >= image.size[1]:
-            asked_height = None
+        :return: self
+        :rtype: ImageProcess
+
+        :raise: ValueError if `verify_resolution` is True and the image is too large
+        :raise: binascii.Error: if the base64 is incorrect
+        :raise: OSError if the image can't be identified by PIL
+        """
+        self.base64_source = base64_source or False
+
+        if not base64_source or base64_source[:1] == b'P':
+            # don't process empty source or SVG
+            self.image = False
         else:
-            asked_width = None
-        if asked_width is None and asked_height is None:
-            return base64_source
+            self.image = base64_to_image(self.base64_source)
 
-    if asked_width is None:
-        asked_width = int(image.size[0] * (float(asked_height) / image.size[1]))
-    if asked_height is None:
-        asked_height = int(image.size[1] * (float(asked_width) / image.size[0]))
-    size = asked_width, asked_height
-    # check image size: do not create a thumbnail if avoiding smaller images
-    if avoid_if_small and image.size[0] <= size[0] and image.size[1] <= size[1]:
-        return base64_source
+            w, h = self.image.size
+            if verify_resolution and w * h > IMAGE_MAX_RESOLUTION:
+                raise ValueError(_("Image size excessive, uploaded images must be smaller than %s million pixels.") % str(IMAGE_MAX_RESOLUTION / 10e6))
 
-    if image.size != size:
-        image = image_resize_and_sharpen(image, size, preserve_aspect_ratio=preserve_aspect_ratio, upper_limit=upper_limit)
-    if image.mode not in ["1", "L", "P", "RGB", "RGBA"] or (filetype == 'JPEG' and image.mode == 'RGBA'):
-        image = image.convert("RGB")
+            self.original_format = self.image.format
 
-    background_stream = io.BytesIO()
-    image.save(background_stream, filetype)
-    return codecs.encode(background_stream.getvalue(), encoding)
+    def image_base64(self, quality=80, output_format=None):
+        """Return the base64 encoded image resulting of all the image processing
+        operations that have been applied previously.
+
+        Return False if the initialized `base64_source` was falsy, and return
+        the initialized `base64_source` without change if it was SVG.
+
+        :param quality: quality setting to apply.
+            - Ignored if image is not JPEG.
+            - 1 is worse, 95 is best. Default to 80.
+        :type quality: int
+
+        :param output_format: the output format. Can be PNG, JPEG, GIF, or ICO.
+            Default to the format of the original image. BMP is converted to
+            PNG, other formats than those mentioned above are converted to JPEG.
+        :type output_format: string
+
+        :return: image base64 encoded or False
+        :rtype: bytes or False
+        """
+        output_image = self.image
+
+        if not output_image:
+            return self.base64_source
+
+        output_format = (output_format or self.original_format).upper()
+        if output_format == 'BMP':
+            output_format = 'PNG'
+        elif output_format not in ['PNG', 'JPEG', 'GIF', 'ICO']:
+            output_format = 'JPEG'
+
+        opt = {'format': output_format}
+
+        if output_format == 'PNG':
+            opt['optimize'] = True
+            alpha = False
+            if output_image.mode in ('RGBA', 'LA') or (output_image.mode == 'P' and 'transparency' in output_image.info):
+                alpha = output_image.convert('RGBA').split()[-1]
+            if output_image.mode != 'P':
+                # Floyd Steinberg dithering by default
+                output_image = output_image.convert('RGBA').convert('P', palette=Image.WEB, colors=256)
+            if alpha:
+                output_image.putalpha(alpha)
+        if output_format == 'JPEG':
+            opt['optimize'] = True
+            opt['quality'] = quality
+        if output_format == 'GIF':
+            opt['optimize'] = True
+
+        if output_image.mode not in ["1", "L", "P", "RGB", "RGBA"] or (output_format == 'JPEG' and output_image.mode == 'RGBA'):
+            output_image = output_image.convert("RGB")
+
+        return image_to_base64(output_image, **opt)
+
+    def resize(self, max_width=0, max_height=0):
+        """Resize the image.
+
+        The image is never resized above the current image size. This method is
+        only to create a smaller version of the image.
+
+        The current ratio is preserved. To change the ratio, see `crop_resize`.
+
+        If `max_width` or `max_height` is falsy, it will be computed from the
+        other to keep the current ratio. If both are falsy, no resize is done.
+
+        :param max_width: max width
+        :type max_width: int
+
+        :param max_height: max height
+        :type max_height: int
+
+        :return: self to allow chaining
+        :rtype: ImageProcess
+        """
+        if self.image and (max_width or max_height):
+            w, h = self.image.size
+            asked_width = max_width or (w * max_height) // h
+            asked_height = max_height or (h * max_width) // w
+            self.image.thumbnail((asked_width, asked_height), Image.LANCZOS)
+        return self
+
+    def crop_resize(self, max_width, max_height, center_x=0.5, center_y=0.5):
+        """Crop and resize the image.
+
+        The image is never resized above the current image size. This method is
+        only to create smaller versions of the image.
+
+        Instead of preserving the ratio of the original image like `resize`,
+        this method will force the output to take the ratio of the given
+        `max_width` and `max_height`, so both have to be defined.
+
+        The crop is done before the resize in order to preserve as much of the
+        original image as possible. The goal of this method is primarily to
+        resize to a given ratio, and it is not to crop unwanted parts of the
+        original image. If the latter is what you want to do, you should create
+        another method, or directly use the `crop` method from PIL.
+
+        :param max_width: max width
+        :type max_width: int
+
+        :param max_height: max height
+        :type max_height: int
+
+        :param center_x: the center of the crop between 0 (left) and 1 (right)
+            Default to 0.5 (center).
+        :type center_x: float
+
+        :param center_y: the center of the crop between 0 (top) and 1 (bottom)
+            Default to 0.5 (center).
+        :type center_y: float
+
+        :return: self to allow chaining
+        :rtype: ImageProcess
+        """
+        if self.image and max_width and max_height:
+            w, h = self.image.size
+            # We want to keep as much of the image as possible -> at least one
+            # of the 2 crop dimensions always has to be the same value as the
+            # original image.
+            # The target size will be reached with the final resize.
+            if w / max_width > h / max_height:
+                new_w, new_h = w, (max_height * w) // max_width
+            else:
+                new_w, new_h = (max_width * h) // max_height, h
+
+            # No cropping above image size.
+            if new_w > w:
+                new_w, new_h = w, (new_h * w) // new_w
+            if new_h > h:
+                new_w, new_h = (new_w * h) // new_h, h
+
+            # Corretly place the center of the crop.
+            x_offset = (w - new_w) * center_x
+            h_offset = (h - new_h) * center_y
+
+            self.image = self.image.crop((x_offset, h_offset, x_offset + new_w, h_offset + new_h))
+
+        return self.resize(max_width, max_height)
+
+    def colorize(self):
+        """Replace the trasparent background by a random color.
+
+        :return: self to allow chaining
+        :rtype: ImageProcess
+        """
+        if self.image:
+            original = self.image
+            color = (randrange(32, 224, 24), randrange(32, 224, 24), randrange(32, 224, 24))
+            self.image = Image.new('RGB', original.size)
+            self.image.paste(color, box=(0, 0) + original.size)
+            self.image.paste(original, mask=original)
+        return self
 
 
-def image_resize_and_sharpen(image, size, preserve_aspect_ratio=False, factor=2.0, upper_limit=False):
+def image_process(base64_source, size=(0, 0), verify_resolution=False, quality=80, crop=None, colorize=False, output_format=None):
+    """Process the `base64_source` image by executing the given operations and
+    return the result as a base64 encoded image.
     """
-        Create a thumbnail by resizing while keeping ratio.
-        A sharpen filter is applied for a better looking result.
-
-        :param image: PIL.Image.Image()
-        :param size: 2-tuple(width, height)
-        :param preserve_aspect_ratio: boolean (default: False)
-        :param factor: Sharpen factor (default: 2.0)
-    """
-    origin_mode = image.mode
-    if image.mode != 'RGBA':
-        image = image.convert('RGBA')
-    image.thumbnail(size, Image.ANTIALIAS)
-    if preserve_aspect_ratio:
-        size = image.size
-    sharpener = ImageEnhance.Sharpness(image)
-    resized_image = sharpener.enhance(factor)
-    # create a transparent image for background and paste the image on it
-    if upper_limit:
-        image = Image.new('RGBA', (size[0], size[1]-3), (255, 255, 255, 0)) # FIXME temporary fix for trimming the ghost border.
-    else:
-        image = Image.new('RGBA', size, (255, 255, 255, 0))
-    image.paste(resized_image, ((size[0] - resized_image.size[0]) // 2, (size[1] - resized_image.size[1]) // 2))
-
-    if image.mode != origin_mode:
-        image = image.convert(origin_mode)
-    return image
-
-
-def image_save_for_web(image, fp=None, format=None):
-    """
-        Save image optimized for web usage.
-
-        :param image: PIL.Image.Image()
-        :param fp: File name or file object. If not specified, a bytestring is returned.
-        :param format: File format if could not be deduced from image.
-    """
-    opt = dict(format=image.format or format)
-    if image.format == 'PNG':
-        opt.update(optimize=True)
-        alpha = False
-        if image.mode in ('RGBA', 'LA') or (image.mode == 'P' and 'transparency' in image.info):
-            alpha = image.convert('RGBA').split()[-1]
-        if image.mode != 'P':
-            # Floyd Steinberg dithering by default
-            image = image.convert('RGBA').convert('P', palette=Image.WEB, colors=256)
-        if alpha:
-            image.putalpha(alpha)
-    elif image.format == 'JPEG':
-        opt.update(optimize=True, quality=80)
-    if fp:
-        image.save(fp, **opt)
-    else:
-        img = io.BytesIO()
-        image.save(img, **opt)
-        return img.getvalue()
-
-
-def image_resize_image_big(base64_source, encoding='base64', filetype=None, avoid_if_small=True, preserve_aspect_ratio=False, upper_limit=False):
-    """ Wrapper on image_resize_image, to resize images larger than the standard
-        'big' image size: 1024x1024px.
-        Refer to image_resize_image for the parameters.
-    """
-    return image_resize_image(base64_source, IMAGE_BIG_SIZE, encoding, filetype, avoid_if_small, preserve_aspect_ratio, upper_limit)
-
-
-def image_resize_image_large(base64_source, encoding='base64', filetype=None, avoid_if_small=True, preserve_aspect_ratio=False, upper_limit=False):
-    """ Wrapper on image_resize_image, to resize to the standard 'large'
-        image size: 256x256.
-        Refer to image_resize_image for the parameters.
-    """
-    return image_resize_image(base64_source, IMAGE_LARGE_SIZE, encoding, filetype, avoid_if_small, preserve_aspect_ratio, upper_limit)
-
-
-def image_resize_image_medium(base64_source, encoding='base64', filetype=None, avoid_if_small=False, preserve_aspect_ratio=False, upper_limit=False):
-    """ Wrapper on image_resize_image, to resize to the standard 'medium'
-        image size: 128x128.
-        Refer to image_resize_image for the parameters.
-    """
-    return image_resize_image(base64_source, IMAGE_MEDIUM_SIZE, encoding, filetype, avoid_if_small, preserve_aspect_ratio, upper_limit)
-
-
-def image_resize_image_small(base64_source, encoding='base64', filetype=None, avoid_if_small=False, preserve_aspect_ratio=False, upper_limit=False):
-    """ Wrapper on image_resize_image, to resize to the standard 'small' image
-        size: 64x64.
-        Refer to image_resize_image for the parameters.
-    """
-    return image_resize_image(base64_source, IMAGE_SMALL_SIZE, encoding, filetype, avoid_if_small, preserve_aspect_ratio, upper_limit)
-
-
-# ----------------------------------------
-# Crop Image
-# ----------------------------------------
-
-def crop_image(data, type='top', ratio=None, size=None, image_format=None):
-    """ Used for cropping image and create thumbnail
-        :param data: base64 data of image.
-        :param type: Used for cropping position possible
-            Possible Values : 'top', 'center', 'bottom'
-        :param ratio: Cropping ratio
-            e.g for (4,3), (16,9), (16,10) etc
-            send ratio(1,1) to generate square image
-        :param size: Resize image to size
-            e.g (200, 200)
-            after crop resize to 200x200 thumbnail
-        :param image_format: return image format PNG,JPEG etc
-    """
-    if not data:
-        return False
-    image_stream = Image.open(io.BytesIO(base64.b64decode(data)))
-    output_stream = io.BytesIO()
-    w, h = image_stream.size
-    new_h = h
-    new_w = w
-
-    if ratio:
-        w_ratio, h_ratio = ratio
-        new_h = (w * h_ratio) // w_ratio
-        new_w = w
-        if new_h > h:
-            new_h = h
-            new_w = (h * w_ratio) // h_ratio
-
-    image_format = image_format or image_stream.format or 'JPEG'
-    if type == "top":
-        cropped_image = image_stream.crop((0, 0, new_w, new_h))
-        cropped_image.save(output_stream, format=image_format)
-    elif type == "center":
-        cropped_image = image_stream.crop(((w - new_w) // 2, (h - new_h) // 2, (w + new_w) // 2, (h + new_h) // 2))
-        cropped_image.save(output_stream, format=image_format)
-    elif type == "bottom":
-        cropped_image = image_stream.crop((0, h - new_h, new_w, h))
-        cropped_image.save(output_stream, format=image_format)
-    else:
-        raise ValueError('ERROR: invalid value for crop_type')
+    image = ImageProcess(base64_source, verify_resolution)
     if size:
-        thumbnail = Image.open(io.BytesIO(output_stream.getvalue()))
-        output_stream.truncate(0)
-        output_stream.seek(0)
-        thumbnail.thumbnail(size, Image.ANTIALIAS)
-        thumbnail.save(output_stream, image_format)
-    return base64.b64encode(output_stream.getvalue())
-
-
-# ----------------------------------------
-# Colors
-# ---------------------------------------
-
-def image_colorize(original, randomize=True, color=(255, 255, 255)):
-    """ Add a color to the transparent background of an image.
-        :param original: file object on the original image file
-        :param randomize: randomize the background color
-        :param color: background-color, if not randomize
-    """
-    # create a new image, based on the original one
-    original = Image.open(io.BytesIO(original))
-    image = Image.new('RGB', original.size)
-    # generate the background color, past it as background
-    if randomize:
-        color = (randrange(32, 224, 24), randrange(32, 224, 24), randrange(32, 224, 24))
-    image.paste(color, box=(0, 0) + original.size)
-    image.paste(original, mask=original)
-    # return the new image
-    buffer = io.BytesIO()
-    image.save(buffer, 'PNG')
-    return buffer.getvalue()
+        if crop:
+            center_x = 0.5
+            center_y = 0.5
+            if crop == 'top':
+                center_y = 0
+            elif crop == 'bottom':
+                center_y = 1
+            image.crop_resize(max_width=size[0], max_height=size[1], center_x=center_x, center_y=center_y)
+        else:
+            image.resize(max_width=size[0], max_height=size[1])
+    if colorize:
+        image.colorize()
+    return image.image_base64(quality=quality, output_format=output_format)
 
 
 # ----------------------------------------
 # Misc image tools
 # ---------------------------------------
 
-def is_image_size_above(base64_source, encoding='base64', size=IMAGE_BIG_SIZE):
+def base64_to_image(base64_source):
+    """Return a PIL image from the given `base64_source`.
+
+    :param base64_source: the image base64 encoded
+    :type base64_source: string or bytes
+
+    :return: the PIL image
+    :rtype: PIL.Image
+
+    :raise: binascii.Error: if the base64 is incorrect
+    :raise: OSError if the image can't be identified by PIL
+    """
+    return Image.open(io.BytesIO(base64.b64decode(base64_source)))
+
+
+def image_to_base64(image, format, **params):
+    """Return a base64_image from the given PIL `image` using `params`.
+
+    :param image: the PIL image
+    :type image: PIL.Image
+
+    :param params: params to expand when calling PIL.Image.save()
+    :type params: dict
+
+    :return: the image base64 encoded
+    :rtype: bytes
+    """
+    stream = io.BytesIO()
+    image.save(stream, format=format, **params)
+    return base64.b64encode(stream.getvalue())
+
+
+def is_image_size_above(base64_source, size=IMAGE_BIG_SIZE):
     """Return whether or not the size of the given image `base64_source` is
     above the provided `size` (tuple: width, height).
     """
     if not base64_source:
         return False
-    if isinstance(base64_source, str):
-        base64_source = base64_source.encode('ascii')
     if base64_source[:1] == b'P':
         # False for SVG
         return False
-    image_stream = io.BytesIO(codecs.decode(base64_source, encoding))
-    image = Image.open(image_stream)
+    image = base64_to_image(base64_source)
     width, height = image.size
     return width > size[0] or height > size[1]
 
 
+def image_guess_size_from_field_name(field_name):
+    """Attempt to guess the image size based on `field_name`.
+
+    If it can't be guessed, return (0, 0) instead.
+
+    :param field_name: the name of a field
+    :type field_name: string
+
+    :return: the guessed size
+    :rtype: tuple (width, height)
+    """
+    suffix = 'big' if field_name == 'image' else field_name.split('_')[-1]
+    if suffix == 'big':
+        return IMAGE_BIG_SIZE
+    if suffix == 'large':
+        return IMAGE_LARGE_SIZE
+    if suffix == 'medium':
+        return IMAGE_MEDIUM_SIZE
+    if suffix == 'small':
+        return IMAGE_SMALL_SIZE
+    return (0, 0)
+
+
 def image_get_resized_images(base64_source,
-        big_name='image', large_name='image_large', medium_name='image_medium', small_name='image_small',
-        avoid_resize_big=True, avoid_resize_large=True, avoid_resize_medium=True, avoid_resize_small=True,
-        preserve_aspect_ratio=False, upper_limit=False):
+        big_name='image', large_name='image_large', medium_name='image_medium', small_name='image_small'):
     """ Standard tool function that returns a dictionary containing the
         big, medium, large and small versions of the source image.
 
@@ -318,23 +334,20 @@ def image_get_resized_images(base64_source,
             previous parameters.
     """
     return_dict = dict()
-    if isinstance(base64_source, str):
-        base64_source = base64_source.encode('ascii')
     if big_name:
-        return_dict[big_name] = image_resize_image_big(base64_source, avoid_if_small=avoid_resize_big, preserve_aspect_ratio=preserve_aspect_ratio, upper_limit=upper_limit)
+        return_dict[big_name] = image_process(base64_source, size=IMAGE_BIG_SIZE)
     if large_name:
-        return_dict[large_name] = image_resize_image_large(base64_source, avoid_if_small=avoid_resize_large, preserve_aspect_ratio=preserve_aspect_ratio, upper_limit=upper_limit)
+        return_dict[large_name] = image_process(base64_source, size=IMAGE_LARGE_SIZE)
     if medium_name:
-        return_dict[medium_name] = image_resize_image_medium(base64_source, avoid_if_small=avoid_resize_medium, preserve_aspect_ratio=preserve_aspect_ratio, upper_limit=upper_limit)
+        return_dict[medium_name] = image_process(base64_source, size=IMAGE_MEDIUM_SIZE)
     if small_name:
-        return_dict[small_name] = image_resize_image_small(base64_source, avoid_if_small=avoid_resize_small, preserve_aspect_ratio=preserve_aspect_ratio, upper_limit=upper_limit)
+        return_dict[small_name] = image_process(base64_source, size=IMAGE_SMALL_SIZE)
     return return_dict
 
 
 def image_resize_images(vals,
         return_big=True, return_large=False, return_medium=True, return_small=True,
-        big_name='image', large_name='image_large', medium_name='image_medium', small_name='image_small',
-        preserve_aspect_ratio=False, upper_limit=False):
+        big_name='image', large_name='image_large', medium_name='image_medium', small_name='image_small'):
     """ Update ``vals`` with image fields resized as expected. """
     big_image = vals.get(big_name)
     large_image = vals.get(large_name)
@@ -345,11 +358,7 @@ def image_resize_images(vals,
 
     if biggest_image:
         vals.update(image_get_resized_images(biggest_image,
-            big_name=return_big and big_name, large_name=return_large and large_name, medium_name=return_medium and medium_name, small_name=return_small and small_name,
-            avoid_resize_big=True, avoid_resize_large=True,
-            avoid_resize_medium=(not big_image and not large_image),
-            avoid_resize_small=(not big_image and not large_image and not medium_image),
-            preserve_aspect_ratio=preserve_aspect_ratio, upper_limit=upper_limit))
+            big_name=return_big and big_name, large_name=return_large and large_name, medium_name=return_medium and medium_name, small_name=return_small and small_name))
     elif any(f in vals for f in [big_name, large_name, medium_name, small_name]):
         if return_big:
             vals[big_name] = False
@@ -361,30 +370,6 @@ def image_resize_images(vals,
             vals[small_name] = False
 
 
-def limited_image_resize(content, width=None, height=None, crop=False, upper_limit=False, avoid_if_small=False):
-    """
-    :param content: bytes (should be an image)
-    """
-    if content and (width or height):
-        signatures = [b'\xFF\xD8\xFF', b'\x89PNG\r\n\x1A\n']
-        decoded_content = base64.b64decode(content)
-        is_image = any(decoded_content.startswith(signature) for signature in signatures)
-        if is_image:
-            height = int(height or 0)
-            width = int(width or 0)
-            if crop:
-                return crop_image(content, type='center', size=(width, height), ratio=(1, 1))
-            else:
-                if not upper_limit:
-                    # resize maximum 500*500
-                    width = min(width, 500)
-                    height = min(height, 500)
-                return image_resize_image(
-                    base64_source=content, size=(width or None, height or None), encoding='base64', upper_limit=upper_limit,
-                    avoid_if_small=avoid_if_small)
-    return content
-
-
 def image_data_uri(base64_source):
     """This returns data URL scheme according RFC 2397
     (https://tools.ietf.org/html/rfc2397) for all kind of supported images
@@ -402,5 +387,5 @@ if __name__=="__main__":
     assert len(sys.argv)==3, 'Usage to Test: image.py SRC.png DEST.png'
 
     img = base64.b64encode(open(sys.argv[1],'rb').read())
-    new = image_resize_image(img, (128,100))
+    new = image_process(img, (128, 100))
     open(sys.argv[2], 'wb').write(base64.b64decode(new))
diff --git a/odoo/tools/mimetypes.py b/odoo/tools/mimetypes.py
index 4350e8c71a5..79bced25c5d 100644
--- a/odoo/tools/mimetypes.py
+++ b/odoo/tools/mimetypes.py
@@ -125,6 +125,7 @@ _mime_mappings = (
     _Entry('image/svg+xml', [b'<'], [
         _check_svg,
     ]),
+    _Entry('image/x-icon', [b'\x00\x00\x01\x00'], []),
     # OLECF files in general (Word, Excel, PPT, default to word because why not?)
     _Entry('application/msword', [b'\xD0\xCF\x11\xE0\xA1\xB1\x1A\xE1', b'\x0D\x44\x4F\x43'], [
         _check_olecf