Secure robust high-fidelity watermarking

• US 7,298,865 B2
• Filed: 04/18/2002
• Issued: 11/20/2007
• Est. Priority Date: 07/30/2001
• Status: Active Grant

• Current Assignee: Transpacific Information, LLC
• Original Assignee: Sarnoff Corporation (SRI International, Inc.)
• Inventors: Bloom, Jeffrey Adam, Lubin, Jeffrey
• Primary Examiner(s): Bali; Vikkram
• Assistant Examiner(s): LAROSE, COLIN M
• Application Number:
  US10/124,995
• Publication Number:
  US 20030021439A1
• Time in Patent Office: 2,042 Days
• Field of Search: 382/100, 358/3.28, 713/176, 380/28, 380/30, 380/54
• US Class Current: 382/100
• CPC Class Codes: G06T 1/0071 using multiple or alternating watermarks
• CPC Class Codes: G06T 1/0085 Time domain based watermarking, e.g. watermarks spread over several images

Abstract

For each small image region (in space and time), a measure of perceptual transparence of each of a set of possible watermark carrier modulations is used to choose a subset of such modulations, from which a secure random number generator selects, for each image region, a single carrier, modulations of which carry the watermark data.

Claims

1. A method for encoding watermark data into an image sequence, comprising:

(a) identifying a set of one or more watermark carriers for potential use in encoding the watermark data into the image sequence;

(b) analyzing content of the image sequence to identify a plurality of potential locations in the image sequence for encoding the watermark data into the image sequence, wherein each potential location corresponds to an acceptable encoding of a portion of the watermark data using a particular one of the watermark carriers in the set, wherein the set comprises at least one watermark carrier based on a spatio-temporal frequency pattern, having a frequency below a peak of a spatio-temporal contrast sensitivity function;

(c) selecting a subset of the potential locations; and

(d) for each selected location, modulating the corresponding watermark carrier by a portion of the watermark data and inserting the resulting modulated watermark carrier into the image sequence at the selected location.

2. The invention of claim 1, wherein the portion of the watermark data corresponds to a single bit of watermark data.

3. The invention of claim 1, wherein:

step (b) identifies two or more different watermark carriers for at least one potential location; and

step (c) selects one of the two or more different watermark carriers for that at least one potential location.

4. The invention of claim 1, wherein:

step (b) divides the image sequence into a plurality of regions and identifies a plurality of potential locations in each region; and

step (c) selects one or more of the potential locations in each region.

5. The invention of claim 4, wherein the image sequence is a video work and each region corresponds to a three-dimensional (3D) spatio-temporal volume of video data in the video work.

6. The invention of claim 5, wherein each region corresponds to a plurality of consecutive video frames in the video work.

7. The invention of claim 1, wherein the image sequence is a video work.

8. The invention of claim 7, wherein at least one of the watermark carriers is based on a three-dimensional spatio-temporal pattern and each corresponding modulated watermark carrier is inserted into the video work in the spatial domain.

9. The invention of claim 8, wherein each corresponding modulated watermark carrier is added to luminance components in the video work.

10. The invention of claim 1, wherein the image sequence comprises an audio work.

11. The invention of claim 1, wherein each potential location corresponds to a location where a single modulated watermark carrier can be inserted imperceptibly, but where insertion of two or more modulated watermark carriers at or near the potential location results in perceptible artifacts.

12. The invention of claim 1, wherein each selected location is separated from all other selected locations by a specified minimum distance.

13. The invention of claim 1, wherein step (b) uses a perceptual model to identify the plurality of potential locations.

14. The invention of claim 13, wherein step (b) implements a JND (Just Noticeable Difference) analysis to identify the plurality of potential locations.

15. The invention of claim 1, wherein:

step (b) performs a cluster analysis to characterize distribution of the potential locations; and

step (c) selects the subset of potential locations based on results from the cluster analysis.

16. The invention of claim 15, wherein the subset of potential locations corresponds to clusters of the potential locations having relatively high density.

17. The invention of claim 1, wherein step (c) randomly selects the subset of potential locations.

18. The invention of claim 17, wherein step (c) randomly selects the subset of potential locations based on a sequence of values generated by a random number generator operating in accordance with a secret watermark key.

19. The invention of claim 1, wherein the at least one watermark carrier has a sensitivity at least an order of magnitude below the sensitivity at the peak.

20. The invention of claim 1, wherein the image sequence is a video work and the at least one watermark carrier is based on a three-dimensional spatio-temporal Gaussian pattern whose amplitude falls off from a central peak in all spatial and temporal directions.

21. A method for encoding watermark data into an image sequence, comprising:

(a) modulating a watermark carrier by a portion of the watermark data, wherein the watermark carrier is based on a spatio-temporal frequency pattern having a frequency below a peak of a spatio-temporal contrast sensitivity function; and

(b) inserting the resulting modulated watermark carrier into the image sequence at a selected location.

22. The invention of claim 21, wherein a watermark carrier has a sensitivity at least an order of magnitude below the sensitivity at the peak.