IP Library › Granted Patent US 12682680

Granted Patent B2

US 12682680 · App. 18/453,881 · Granted Jul 14, 2026

Attribute-consistent image fusion and model training methods, apparatuses, and computer devices

Inventors: Ang Li (Hangzhou, CN); Jun Zhou (Hangzhou, CN); Chilin Fu (Hangzhou, CN); Xiaolu Zhang (Hangzhou, CN)

Assignee: Alipay (Hangzhou) Information Technology Co., Ltd.

G06V40/172 G06V40/168 G06T2207/20221

View Patent ↗

Loading inventors, assignments & file history…

Monitor This Case

Get email alerts when status or documents change.

Order Certified Copies

Most orders are placed with the USPTO same day — all within 24 business hours.

Order via The Patent Place →

Pre-filled with this patent's details

Quick Facts

Patent No.: US 12682680
App. No.: 18/453,881
Granted: Jul 14, 2026
Kind: B2

External Links

USPTO Patent Center↗ Full Patent PDF↗

Abstract

A computer-implemented method for image processing includes obtaining, based on a source image including a first biological object, an identity feature and attribute feature used to represent identity information and attribute information of the first biological object. Based on a target image including a second biological object, a context feature used to represent context information of the second biological object is obtained. Based on the identity feature, a fused image of the source image and the target image, the attribute feature, and the context feature is generated, where the fused image includes a fused biological object of a first biological object and the second biological object, identity information and attribute information of the fused biological object is identical to the identity information and the attribute information of the first biological object, and context information of the fused biological object is identical to the context information of the second biological object.

Claims (88)

1 . A computer-implemented method for image processing, comprising:

obtaining, based on a source image, an identity feature and an attribute feature, wherein the source image includes a first biological object, the identity feature is used to represent identity information of the first biological object, and the attribute feature is used to represent attribute information of the first biological object;

obtaining, based on a target image, a context feature, wherein the target image includes a second biological object, and wherein the context feature is used to represent context information of the second biological object; and

generating, based on the identity feature, the attribute feature, and the context feature, a fused image of the source image and the target image, wherein the fused image includes a fused biological object of the first biological object and the second biological object, identity information and attribute information of the fused biological object is identical to the identity information and the attribute information of the first biological object, and context information of the fused biological object is identical to the context information of the second biological object, and wherein the context feature comprises at least one of a posture feature used to represent posture information and a background feature used to represent background information, and the attribute feature includes at least one of:

a beard feature used to represent beard information;

a pouch feature used to represent pouch information;

a gender feature used to represent gender information; and

an age feature used to represent age information.

2 . The computer-implemented method of claim 1 , wherein:

obtaining, based on a source image, an identity feature and an attribute feature, comprises:

inputting the source image into an identity recognition model to obtain the identity feature;

inputting the source image into an attribute recognition model to obtain the attribute feature;

obtaining, based on a target image, a context feature, comprises:

inputting the target image into a context recognition model to obtain the context feature; and

generating a fused image of the source image and the target image, comprises:

inputting the identity feature, the attribute feature, and the context feature into a feature fusion model to obtain the fused image.

3 . The computer-implemented method of claim 1 , comprising:

performing resolution enhancement processing on the source image and the target image; and

obtaining, based on a source image, an identity feature and an attribute feature, comprises:

obtaining, based on a resolution-enhanced source image, the identity feature and the attribute feature; and

obtaining, based on a target image, a context feature, comprises:

obtaining, based on a resolution-enhanced target image, the context feature.

4 . The computer-implemented method of claim 3 , wherein generating a fused image of the source image and the target image, comprises:

generating, based on the identity feature, the fused image, the attribute feature, the context feature, and an intermediate feature, wherein the intermediate feature is a feature extracted based on the source image in the resolution enhancement processing and is used to generate the resolution-enhanced source image.

5 . The computer-implemented method of claim 3 , wherein performing resolution enhancement processing on the source image and the target image, comprises:

inputting the source image and the target image into a resolution enhancement model to obtain the resolution-enhanced source image and the resolution-enhanced target image.

6 . The computer-implemented method of claim 1 , wherein:

the source image and the target image are face images;

the first biological object, the second biological object, and the fused biological object are faces; and

the fused image is a face-shifted image.

7 . A non-transitory, computer-readable medium storing one or more instructions executable by a computer system to perform operations for image processing, comprising:

generating, based on the identity feature, a fused image of the source image and the target image, the attribute feature, and the context feature, wherein the fused image includes a fused biological object of the first biological object and the second biological object, identity information and attribute information of the fused biological object is identical to the identity information and the attribute information of the first biological object, and context information of the fused biological object is identical to the context information of the second biological object, and wherein the context feature comprises at least one of a posture feature used to represent posture information and a background feature used to represent background information, and the attribute feature includes at least one of:

a beard feature used to represent beard information;

a pouch feature used to represent pouch information;

a gender feature used to represent gender information; and

an age feature used to represent age information.

8 . The non-transitory, computer-readable medium of claim 7 , wherein:

obtaining, based on a source image, an identity feature and an attribute feature, comprises:

inputting the source image into an identity recognition model to obtain the identity feature;

inputting the source image into an attribute recognition model to obtain the attribute feature;

obtaining, based on a target image, a context feature, comprises:

inputting the target image into a context recognition model to obtain the context feature; and

generating a fused image of the source image and the target image, comprises:

inputting the identity feature, the attribute feature, and the context feature into a feature fusion model to obtain the fused image.

9 . The non-transitory, computer-readable medium of claim 7 , comprising one or more instructions for:

performing resolution enhancement processing on the source image and the target image; and

obtaining, based on a source image, an identity feature and an attribute feature, comprises:

obtaining, based on a resolution-enhanced source image, the identity feature and the attribute feature; and

obtaining, based on a target image, a context feature, comprises:

obtaining, based on a resolution-enhanced target image, the context feature.

10 . The non-transitory, computer-readable medium of claim 9 , wherein generating a fused image of the source image and the target image, comprises:

11 . The non-transitory, computer-readable medium of claim 9 , wherein performing resolution enhancement processing on the source image and the target image, comprises:

inputting the source image and the target image into a resolution enhancement model to obtain the resolution-enhanced source image and the resolution-enhanced target image.

12 . The non-transitory, computer-readable medium of claim 7 , wherein:

the source image and the target image are face images;

the first biological object, the second biological object, and the fused biological object are faces; and

the fused image is a face-shifted image.

13 . A computer-implemented system, comprising:

one or more computers; and

one or more computer memory devices interoperably coupled with the one or more computers and having tangible, non-transitory, machine-readable media storing one or more instructions that, when executed by the one or more computers, perform one or more operations for image processing, comprising:

a beard feature used to represent beard information;

a pouch feature used to represent pouch information;

a gender feature used to represent gender information; and

an age feature used to represent age information.

14 . The computer-implemented system of claim 13 , wherein:

obtaining, based on a source image, an identity feature and an attribute feature, comprises:

inputting the source image into an identity recognition model to obtain the identity feature;

inputting the source image into an attribute recognition model to obtain the attribute feature;

obtaining, based on a target image, a context feature, comprises:

inputting the target image into a context recognition model to obtain the context feature; and

generating a fused image of the source image and the target image, comprises:

inputting the identity feature, the attribute feature, and the context feature into a feature fusion model to obtain the fused image.

15 . The computer-implemented system of claim 13 , comprising one or more instructions for:

performing resolution enhancement processing on the source image and the target image; and

obtaining, based on a source image, an identity feature and an attribute feature, comprises:

obtaining, based on a resolution-enhanced source image, the identity feature and the attribute feature; and

obtaining, based on a target image, a context feature, comprises:

obtaining, based on a resolution-enhanced target image, the context feature.

16 . The computer-implemented system of claim 15 , wherein generating a fused image of the source image and the target image, comprises:

17 . The computer-implemented system of claim 15 , wherein performing resolution enhancement processing on the source image and the target image, comprises:

inputting the source image and the target image into a resolution enhancement model to obtain the resolution-enhanced source image and the resolution-enhanced target image.