Granted Patent B1

US 12,217,476 · App. 17/363,704 · Granted Feb 4, 2025

Inventor: Alexey Novikov (Quebec, CA)

Assignee: Jumio Corporation

G06V10/507G06T7/97

Abstract

Claims (45)

1. A computer implemented method comprising:

receiving a first image;

generating a first image attribute distribution describing a distribution of values that are based on a first visual attribute associated with the first image, wherein the first visual attribute is at least partially defined by a plurality of pixel values associated with a first set of pixels comprising the first image;

determining, based on the first image attribute distribution, whether the first image is synthetically generated;

determining, based on the first image attribute distribution, that the first image is synthetically generated;

obtaining, responsive to determining that the first image is synthetically generated, a second image, wherein the second image represents a component within the first image;

generating, based on the second image, a second image attribute distribution of pixel values that are based on the first visual attribute associated with the component represented in the second image;

determining, based on the second image attribute distribution, whether the second image is synthetically generated; and

issuing, responsive to determining that the first image is synthetically generated, a rejection; and

identifying, responsive to determining that the second image is synthetically generated, the component as a synthetically component.

2. The computer implemented method of claim 1 , wherein the first visual attribute includes contrast, wherein the first image attribute distribution includes a mean-subtracted contrast normalized coefficient distribution, and wherein determining whether the first image is synthetically generated includes applying a normality test to determine whether the mean-subtracted contrast normalized coefficient distribution is normal.

3. The computer implemented method of claim 1 , wherein the first visual attribute includes a color, wherein the first image attribute distribution includes a color histogram, and wherein determining whether the first image is synthetically generated applies a machine learning algorithm to determine whether an anomaly is present in the color histogram.

4. The computer implemented method of claim 1 , wherein determining whether the first image is synthetically generated, includes applying a machine learning algorithm.

5. The computer implemented method of claim 1 , wherein determining whether the first image is synthetically generated, includes determining whether the first image is partially synthetic or fully synthetic.

6. The computer implemented method of claim 1 further comprising:

receiving a third image, wherein the third image is another component of the first image distinct from the component represented in the second image;

generating, based on the third image, a third image attribute distribution of pixel values that are based on the first visual attribute associated with the another component represented in the third image and the first image; and

determining, based on the third image attribute distribution, whether the another component represented in the third image is a synthetically generated component.

7. The computer implemented method of claim 1 , wherein the first image is a first image component of a larger image.

8. The computer implemented method of claim 1 , wherein the first image is a component of a larger image, and the first image includes an area of interest within the larger image and a margin beyond the area of interest in one or more dimensions.

9. The computer implemented method of claim 1 , wherein the rejection triggers one or more of a failure of identity confirmation and failure of authentication.

10. The method of claim 1 , wherein the first image includes one or more of a face and text associated with a field in a document.

11. A system comprising:

a processor; and

a memory, the memory storing instructions that, when executed by the processor, cause the system to:

receive a first image;

generate a first image attribute distribution describing a distribution of values that are based on a first visual attribute associated with the first image, wherein the first visual attribute is at least partially defined by a plurality of pixel values associated with a first set of pixels comprising the first image;

determine, based on the first image attribute distribution, whether the first image is synthetically generated;

determine, based on the first image attribute distribution, that the first image is synthetically generated;

obtain, responsive to determining that the first image is synthetically generated, a second image, wherein the second image represents a component within the first image;

generate, based on the second image, a second image attribute distribution of pixel values that are based on the first visual attribute associated with the component represented in the second image; and

determine, based on the second image attribute distribution, whether the second image is synthetically generated; and

issue, responsive to determining that the first image is synthetically generated, a rejection.

12. The system of claim 11 , wherein the first visual attribute includes contrast, wherein the first image attribute distribution includes a mean-subtracted contrast normalized coefficient distribution, and wherein determining whether the first image is synthetically generated includes applying a normality test to determine whether the mean-subtracted contrast normalized coefficient distribution is normal.

13. The system of claim 11 , wherein the first visual attribute includes a color, wherein the first image attribute distribution includes a color histogram, and wherein determining whether the first image is synthetically generated applies a machine learning algorithm to determine whether an anomaly is present in the color histogram.

14. The system of claim 11 , wherein determining whether the first image is synthetically generated, includes applying a machine learning algorithm.

15. The system of claim 11 , wherein determining whether the first image is synthetically generated, includes determining whether the first image is partially synthetic or fully synthetic.

16. The system of claim 11 , the memory further storing instructions that, when executed by the processor, cause the system to:

receive a third image, wherein the third image is another component of the first image distinct from the component represented in the second image;

generate, based on the third image, a third image attribute distribution of pixel values that are based on the first visual attribute associated with the another component represented in the third image and the first image; and

determine, based on the third image attribute distribution, whether the another component represented in the third image is a synthetically generated component.

17. The system of claim 11 , wherein the first image is a first image component of a larger image.

18. The system of claim 11 , wherein the first image is a component of a larger image, and the first image includes an area of interest within the larger image and a margin beyond the area of interest in one or more dimensions.

19. The system of claim 11 , wherein the rejection triggers one or more of a failure of identity confirmation and failure of authentication.

20. The system of claim 11 , wherein the first image includes one or more of a face and text associated with a field in a document.

Assignments (1)

References Cited (14)