Granted Patent B2

US 12,597,095 · App. 18/179,540 · Granted Apr 7, 2026

Inventor: Mrinal Kumar Sharma (Jharkhand, IN)

Assignee: Microsoft Technology Licensing, LLC

G06T5/50G06T7/0002G06V10/761G06T2207/20081G06T2207/20084G06T2207/30168

Abstract

Claims (47)

1 . A data processing system comprising:

a processor; and

a memory in communication with the processor, the memory comprising executable instructions that, when executed by the processor, cause the data processing system to perform functions of:

identifying a genre for an input image using a genre identification machine-learning model;

identifying one or more objects in the input image using an object identification machine-learning model;

comparing the identified genre and the identified one or more objects to a list of genre and object tags for images in an image library to identify a plurality of genre and object tags that are similar to the identified genre and the identified one or more objects, the tags for the images in the image library indicating edits to the images of the image library, the edits comprising at least two of: image filtering, resizing, changing color characteristics, adding text, removing text, adding borders and removing borders;

identifying a list of edits corresponding to each of the identified plurality of genre and object tags;

applying the list of edits to the input image to generate a plurality of enhanced images for the input image;

measuring an aesthetic value for one or more of the plurality of enhanced images; and

recommending, based on the aesthetic value, at least one of the one or more of the plurality of enhanced images as a recommendation for an enhanced version of the input image,

wherein, when new editing features are made available as enhancements, the list of edits is automatically updated using a self-learning mechanism to enable use of the new editing features in enhancing the input image.

2 . The data processing system of claim 1 , wherein the memory further comprises executable instructions that, when executed by the processor, cause the data processing system to perform functions of measuring an aesthetic value for the input image.

3 . The data processing system of claim 2 , wherein the aesthetic value for the input image is compared against the aesthetic value for the enhanced images and enhanced images having an aesthetic value that exceeds the aesthetic value of the input image are provided as the recommendation.

4 . The data processing system of claim 1 , wherein recommending, based on the aesthetic value, at least one of the one or more of the plurality of enhanced images as the recommendation includes recommending the enhanced image, when the aesthetic value of the enhanced image exceeds a threshold value.

5 . The data processing system of claim 1 , wherein the aesthetic value is used in ranking the enhanced images.

6 . The data processing system of claim 5 , wherein the ranking is used in ordering the enhanced images when recommendations are presented to a user.

7 . The data processing system of claim 1 , wherein self-learning mechanism includes identifying genre-object tags to which a new editing feature applies and adding the new editing feature as an edit to the list of edits for the genre-object tags.

8 . The data processing system of claim 1 , wherein a user can select a recommended enhanced image to automatically transform the input image into the enhanced image.

9 . A method for automatically enhancing an input image comprising:

receiving a request over a communication network to enhance the input image;

identifying a genre for the input image via a genre identification machine-learning model;

identifying one or more objects in the input image via an object identification machine-learning model;

comparing the identified genre and the identified one or more objects to a list of genre-object tags for images in a library of genre-object tags associated with enhanced images to identify a plurality of genre-object tags that are similar to the identified genre and the identified one or more objects, the tags for the images in the image library indicating edits to the images of the image library, the edits comprising at least two of:

image filtering, resizing, changing color characteristics, adding text, removing text, adding borders and removing borders;

applying one or more edits associated with the plurality of genre-objects tags to the input image to generate a plurality of enhanced images for the input image;

measuring an aesthetic value for one or more of the plurality of enhanced images; and

recommending, based on the aesthetic value, at least one of the one or more of the plurality of enhanced images as a recommendation for an enhanced version of the input image,

wherein, when new editing features are made available as enhancements, a list of edits is automatically updated using a self-learning mechanism to enable use of the new editing features in enhancing the input image.

10 . The method of claim 9 , further comprising measuring an aesthetic value for the input image.

11 . The method of claim 10 , wherein the aesthetic value for the input image is compared against the aesthetic value for the enhanced images and only enhanced images having an aesthetic value that is higher than the aesthetic value of the input image are provided as the recommendation.

12 . The method of claim 9 , wherein recommending, based on the aesthetic value, at least one of the one or more of the plurality of enhanced images as the recommendation for enhancing the input image includes recommending the enhanced image, when the aesthetic value of the enhanced image exceeds a threshold value.

13 . The method of claim 9 , wherein the aesthetic value is used in ranking the enhanced images.

14 . The method of claim 13 , wherein the ranking is used in ordering the enhanced images when recommendations are presented to a user.

15 . The method of claim 9 , wherein self-learning mechanism includes identifying genre-object tags to which a new editing feature applies and adding the new editing feature as an edit to the list of edits for the genre-object tags.

16 . The method of claim 9 , wherein a user can select a recommended enhanced image to automatically enhance the input image in accordance with the selected enhanced image.

17 . A non-transitory computer readable medium on which are stored instructions that, when executed, cause a programmable device to perform functions of:

accessing a library of images, each image in the library having an associated enhanced version of the image;

identifying a genre for each of a plurality of the images using a genre identification machine-learning model;

identifying one or more objects in each of the plurality of the images using an object identification machine-learning model;

using the identified genres and the identified one or more objects to cluster the images into a plurality of genre-object tags, the tags for the images in the library of images indicating edits to the images of the library, the edits comprising at least two of: image filtering, resizing, changing color characteristics, adding text, removing text, adding borders and removing borders;

identifying a list of edits that were applied to the images in each of the plurality of genre-object tags;

identifying one or more edits corresponding to each of the identified plurality of genre and object tags;

creating, based on the identified one or more edits, a list of edits for each of the plurality of genre-object tags; and

utilizing the list of edits in automatically enhancing an input image.

18 . The non-transitory computer readable medium of claim 17 , wherein when new editing features are made available as enhancements, the list of edits is automatically updated using a self-learning mechanism to enable use of the new editing features in enhancing the input image.

19 . The non-transitory computer readable medium of claim 18 , wherein the self-learning mechanism includes identifying genre-object tags to which a new editing feature applies and adding the new editing feature as an edit to the list of edits for the genre-object tags.

20 . The non-transitory computer readable medium of claim 17 , wherein a filtering unit is utilized to filter out images in the library of images that have a similarity value that exceeds a threshold.

Assignments (2)

Continuity (1)

References Cited (20)