Granted Patent B2

US 12,633,145 · App. 18/143,306 · Granted May 19, 2026

Inventors: Michael Scott Brown (Toronto, CA); Abhijith Punnappurath (North York, CA); Abdelrahman Abdelhamed (Scarborough, CA); Luxi Zhao (Toronto, CA)

Assignee: SAMSUNG ELECTRONICS CO., LTD.

G06V20/95G06V10/82

Abstract

Claims (65)

1 . An electronic device for performing image authentication on an image that is captured by an artificial intelligence (AI) camera module or processed by an AI-image signal processor (ISP), the electronic device comprising:

at least memory storing instructions; and

at least one processor configured to execute the instructions to:

obtain an input image;

obtain processed images of the input image via an AI-based model comprising a plurality of neural networks;

identify at least one neural network of the plurality of neural networks that generates fake pixels, based on a loss type of each of the plurality of neural networks;

for at least one of the processed images that is output from the identified at least one neural network, generate authentication metadata that indicates the fake pixels that have been generated by the AI-based model;

store the authentication metadata in association with the at least one of the processed images in the at least one memory; and

output the at least one of the processed images with an indication of the fake pixels.

2 . The electronic device of claim 1 , wherein the authentication metadata is a binary authentication mask that indicates the fake pixels which are generated by adding, deleting, or modifying at least one scene content included in the input image.

3 . The electronic device of claim 1 ,

the at least one processor is further configured to: based on each of the plurality of neural networks generating the fake pixels, generate a plurality of authentication metadata for the processed images that are output from the plurality of neural networks, respectively.

4 . The electronic device of claim 1 , wherein

the at least one processor is further configured to:

skip generation of the authentication metadata for at least one remaining neural network that is identified as not generating the fake pixels, among the plurality of neural networks.

5 . The electronic device of claim 4 , wherein the at least one processor is further configured to:

identify the at least one neural network that is trained using a perceptual loss or an adversarial loss, as the at least one neural network that generates the fake pixels; and

identify a neural network that is trained using a reconstruction loss, or is configured to change colors of the input image without altering scene content, as the at least one remaining neural network that does not generate the fake pixels.

6 . The electronic device of claim 1 , wherein the at least one processor is further configured to:

generate the authentication metadata by:

applying to the input image a color transform that maps colors of the input image to an output image of at least one neural network of the AI-based model, to obtain a color transformed image; and

computing a difference metric between the color transformed image and the output image for each pixel.

7 . The electronic device of claim 1 , wherein the at least one processor is further configured to:

generate the authentication metadata via the AI-based model that comprises a generator neural network.

8 . The electronic device of claim 7 , wherein the generator neural network comprises:

a first neural network that is trained using a reconstruction loss at a first training step; and

a second neural network that is jointly trained with a discriminator neural network using an adversarial loss, at a second training step at which network parameters of the first neural network are fixed, and

wherein the at least one processor is further configured to generate the authentication metadata based on an output of the second neural network.

9 . A method for performing image authentication on an image that is captured by an artificial intelligence (AI) camera module or processed by an AI-image signal processor (ISP), the method comprising:

obtaining an input image;

obtaining a processed image of the input image via an artificial intelligence (AI)-based model comprising a plurality of neural networks;

identifying at least one neural network of the plurality of neural networks that generates fake pixels, based on a loss type of each of the plurality of neural networks;

for at least one of the processed images that is output from the identified at least one neural network, generating authentication metadata that indicates the fake pixels that have been generated by the AI-based model;

storing the authentication metadata in association with the at least one of the processed images in the at least one memory; and

outputting the at least one of the processed images with an indication of the fake pixels.

10 . The method of claim 9 , wherein the authentication metadata is a binary authentication mask that indicates the fake pixels which are generated by adding, deleting, or modifying at least one scene content included in the input image.

11 . The method of claim 9 ,

the method further comprises:

based on each of the plurality of neural networks generating the fake pixels, generating a plurality of authentication metadata for the processed images that are output from the plurality of neural networks, respectively.

12 . The method of claim 9 , wherein

the method further comprises:

skipping generation of the authentication metadata for at least one remaining neural network that is identified as not generating the fake pixels, among the plurality of neural networks.

13 . The method of claim 12 , further comprising:

identifying the at least one neural network that is trained using a perceptual loss or an adversarial loss, as the at least one neural network that generates the fake pixels; and

identifying a neural network that is trained using a reconstruction loss, or is configured to change colors of the input images without altering scene content, as the at least one remaining neural network that does not generate the fake pixels.

14 . The method of claim 9 , wherein the generating of the authentication metadata comprises:

applying to the input image a color transform that maps colors of the input image to an output image of at least one neural network of the AI-based model, to obtain a color transformed image;

computing a difference metric between the color transformed image and the output image for each pixel; and

generating the authentication metadata based on the difference metric.

15 . The method of claim 9 , wherein the generating of the authentication metadata comprises:

generating the authentication metadata via the AI-based model that comprises a generator neural network.

16 . The method of claim 15 , wherein the generator neural network comprises:

a first neural network that is trained using a reconstruction loss at a first training step; and

a second neural network that is jointly trained with a discriminator neural network using an adversarial loss, at a second training step at which network parameters of the first neural network are fixed, and

wherein the generating the authentication metadata further comprises: generating the authentication metadata based on an output of the second neural network.

17 . A non-transitory computer readable storage medium that stores instructions to be executed by at least one processor to perform a method for performing image authentication on an image that is captured by an artificial intelligence (AI) camera module or processed by an AI-image signal processor (ISP), the method comprising:

obtaining an input image;

obtaining a processed image of the input image via an artificial intelligence (AI)-based model comprising a plurality of neural networks;

identifying at least one neural network of the plurality of neural networks that generates fake pixels, based on a loss type of each of the plurality of neural networks;

for at least one of the processed images that is output from the identified at least one neural network, generating authentication metadata that indicates the fake pixels that have been generated by the AI-based model;

storing the authentication metadata in association with the at least one of the processed images in the at least one memory; and

outputting the at least one of the processed images with an indication of the fake pixels.

18 . The non-transitory computer readable storage medium of claim 17 , wherein the authentication metadata is a binary authentication mask that indicates the fake pixels which are generated by adding, deleting, or modifying at least one scene content included in the input image.

19 . The non-transitory computer readable storage medium of claim 18 , wherein the outputting the processed image with the indication of the fake pixels, comprises:

displaying the binary authentication mask to be overlaid on the processed image.

Assignments (1)

Continuity (2)

References Cited (24)