Detecting compositing in a previously compressed image

1. A method implemented by a computer system, the method for detecting digital image compositing, said method comprising the steps of:

finding an uncompressed digital image as a portion of a larger digital image having a higher likelihood of exhibiting forgery as compared to other portions of the larger digital image, the portion being found using a region-of-interest detector;

detecting a first artifact in the uncompressed digital image, said first artifact being characteristic of an earlier compression of said uncompressed digital image with a predetermined compression algorithm;

evaluating an interpolation marker in said uncompressed digital image to provide interpolation results;

during said evaluating, counteracting a second artifact of said compression algorithm;

classifying said larger digital image responsive to said interpolation results; and

storing results of the classifying in a computer-readable memory device.

2. The method of claim 1 wherein said second artifact and said interpolation marker are different periodicities in said image.

3. The method of claim 2 wherein said first artifact is a blocking artifact.

4. The method of claim 1 wherein said first artifact is a blocking artifact.

5. The method of claim 1 wherein said method further comprises decompressing said portion of the larger digital image to generate the uncompressed digital image.

6. The method of claim 1 wherein said region of interest detector is selected from the group consisting of: face detectors, main subject detectors, skin region detectors, sky detectors, other object detectors, and combinations of two of more of said detectors of said group.

7. The method of claim 1 further comprising repeating said finding, detecting, evaluating, and counteracting steps with an additional digital image.

8. The method of claim 7 wherein said uncompressed digital image and said additional digital image are each different parts of the larger digital image; and wherein said classifying further comprises classifying said larger digital image responsive to said interpolation results.

9. The method of claim 8 further comprising accepting user input designating at least said uncompressed digital image or said additional digital image.

10. The method of claim 7 wherein said uncompressed digital image and said additional digital image are each a frame, or a portion of a frame, of a video sequence.

11. The method of claim 1 wherein said evaluating further comprises:

generating a one-dimensional periodicity signal from the uncompressed digital image;

computing a Discrete Fourier Transform of said periodicity signal; and

detecting a first set of peaks in the said Discrete Fourier transform; and

wherein said counteracting further comprises excluding from said detecting a second set of peaks in said Discrete Fourier Transform.

12. A computer program product for detecting digital image compositing, the computer program product comprising computer readable storage medium having a computer program stored thereon for performing the steps of claim 1 .

13. A system for detecting digital image compositing, said method comprising the steps of:

means for finding an uncompressed digital image as a portion of a larger digital image having a higher likelihood of exhibiting forgery as compared to other portions of the larger digital image;

means for detecting a first artifact in said uncompressed digital image, said first artifact being characteristic of an earlier compression of said uncompressed digital image with a predetermined compression algorithm;

means for evaluating an interpolation marker in said uncompressed digital image to provide interpolation results;

means for counteracting a second artifact of said compression algorithm during said evaluating; and

means for classifying said larger digital image responsive to said interpolation results.

14. A method implemented by a computer system, the method for detecting digital image compositing, said method comprising the steps of:

detecting the presence of a first artifact in said an uncompressed digital image, said first artifact being characteristic of an earlier compression of said digital image with a predetermined compression algorithm;

finding a plurality of different regions of said uncompressed digital image, at least one of the plurality of different regions having a higher likelihood of exhibiting forgery as compared to another of the plurality of different regions of the uncompressed digital image, at least some of the plurality of different regions being found using a region-of-interest detector;

evaluating an interpolation marker in the different regions to provide interpolation results of each of said regions;

when said first artifact is present in said digital image, counteracting a second artifact of said compression algorithm during said evaluating;

classifying said image responsive to said interpolation results; and

storing results of the classifying in a computer-readable memory device.

15. The method of claim 14 wherein said first artifact is a blocking artifact and said second artifact and said interpolation marker are different periodicities in said image.

16. The method of claim 14 wherein said classifying further comprises comparing interpolation results of at least some of said regions.

17. The method of claim 14 wherein said evaluating of each of said regions further comprises:

generating a one-dimensional periodicity signal from the corresponding region;

computing a Discrete Fourier Transform of said periodicity signal; and

detecting a first set of peaks in the said Discrete Fourier transform; and

wherein said counteracting further comprises excluding from said detecting a second set of peaks in said Discrete Fourier Transform.

18. A method implemented by a computer system, the method for detecting digital image compositing, said method comprising the steps of:

finding an uncompressed digital image as a portion of a larger digital image having a higher likelihood of exhibiting forgery as compared to other portions of the larger digital image, the portion being found using a region-of-interest detector;

detecting a blocking artifact in the uncompressed digital image;

identifying a compression-induced periodicity artifact associated with said blocking artifact;

evaluating periodicity in said uncompressed digital image to provide interpolation results;

during said evaluating, counteracting said compression-induced periodicity artifact;

classifying said larger digital image responsive to said interpolation results; and

storing results of the classifying in a computer-readable memory device.

19. The method of claim 8 , wherein said classifying further comprises comparing interpolation results of said uncompressed digital image and said additional digital image.

20. The method of claim 19 , wherein said additional digital image has a lower likelihood of exhibiting forgery as compared to the uncompressed digital image portion of the larger digital image.

21. The system of claim 13 , wherein the finding, detecting, evaluating, and counteracting are performed for an additional portion of the larger digital image, and wherein the classifying further comprises comparing interpolation results of each of the uncompressed digital image and the additional portion of the larger digital image.

22. The method of claim 18 wherein the finding, detecting, evaluating, and counteracting steps are performed for an additional portion of the larger digital image, and wherein the classifying further comprises comparing interpolation results of each of the uncompressed digital image and the additional portion of the larger digital image.