Full-color visibility model using CSF which varies spatially with local luminance
The present disclosure relate generally to color science and digital watermarking. A full color visibility model has been developed which has good correlation to subjective visibility tests for color patches degraded with a watermark. A relatively better correlation can be achieved with a model that applies a luminance correction to the contrast sensitivity functions (CSF). The model can be applied during the watermark embed process, using a pyramid based method, to obtain equal visibility. Better robustness and visibility can be obtained with equal visibility embed than uniform strength embed. Of course, other features, combinations and claims are disclosed as well.
1. A digital watermarking method comprising:
receiving color image data;
using one or more programmed processors: i) degrading luminance data representing the color image data with a first contrast sensitivity function (CSF1) to predict visibility changes attributable to an application of digital watermarking to the color image data, in which the CSF1 varies depending on luminance values associated with local regions of the color image data, and in which said degrading luminance data yields first results for image areas within the color image data, and ii) degrading chrominance data representing the color image data with a second contrast sensitivity function (CSF2), in which the CSF2 is separate from the CSF1, to predict visibility changes attributable to an application of digital watermarking to the color image data, and in which said degrading chrominance data yields second results for image areas within the color image data; and
using one or more programmed processors, transforming the color image data by digital watermarking plural-bit information therein, in which signal embedding strength of the digital watermarking within different image areas varies based on the first results and the second results.
2. The method of claim 1 in which the luminance data degrading of the data representing the color image data with the CSF1 introduces image blurring, and in which the CSF1 varies so that relatively more blurring occurs as luminance of a local image region decreases.
3. The method of claim 1 in which varying embedding strength across different image regions of the color image data yields uniform predicted visibility of the digital watermarking across the color image data.
4. The method of claim 1 in which the CSF1 varies spatially.
5. The method of claim 4 in which the CSF1 varies spatially in terms of spatial width.
6. The method of claim 1 in which the second contrast sensitivity function (CSF2) comprises a blue-yellow contrast sensitivity function and a red-green contrast sensitivity function.
7. The method of claim 1 in which the CSF2 varies spatially depending on luminance values associated with local regions of the obtained color image data.
8. An apparatus comprising:
an input to obtain color image data;
memory storing a visibility model comprising a luminance contrast sensitivity function (CSF1) and a chrominance contrast sensitivity function (CSF2), in which CSF1 and CSF2 are separate contrast sensitivity functions; and
one or more processors configured for:
processing data representing obtained color image data with the CSF1 and the CSF2 to predict degradation of image areas associated with an application of digital watermarking, in which the CSF1 varies depending on luminance values associated with local regions of the obtained color image data and in which the CSF1 is used for processing luminance data and the CSF2 is used for processing chrominance data;
transforming the obtained color image data with digital watermarking, in which the digital watermarking is guided based on results obtained from the processing including predicted degradation of image areas.
9. The apparatus of claim 8 in which the CSF1 varies spatially.
10. The apparatus of claim 9 in which the CSF1 varies spatially in terms of spatial width.
11. The apparatus of claim 9 in which the CSF2 varies spatially in terms of spatial width.
12. The apparatus of claim 8 in which the processing of the obtained color image data with the CSF1 produces image blurring as the predicted degradation, in which the CSF1 varies so that relatively more blurring occurs as luminance of a local image region decreases.
13. The apparatus of claim 8 in which the digital watermarking is guided based on results obtained from the processing by varying embedding strength across different image areas of the obtained color image data based on predicted degradation of the different image areas.
14. The apparatus of claim 13 in which predicted visibility of the digital watermarking across the different image areas comprises uniform predicted visibility.
15. The apparatus of claim 8 in which said one or more processors are configured for processing the obtained color image data with an attention model to predict visual traffic areas.
16. The apparatus of claim 15 in which the digital watermarking is guided based on the results obtained from the processing including predicted degradation of image areas and on predictions of visual traffic areas obtained from processing the obtained color image data with the attention model.
17. The apparatus of claim 8 in which the chrominance contrast sensitivity function (CSF2) comprises a blue-yellow contrast sensitivity function and a red-green contrast sensitivity function.
18. The apparatus of claim 8 in which the CSF2 varies depending on luminance values associated with local regions of the obtained color image data.
19. An method comprising:
obtaining data representing a color image;
using one or more programmed processors, processing the data representing the color image with a luminance contrast sensitivity function (CSF1) and a chrominance contrast sensitivity function (CSF2) to predict degradation of image areas associated with an application of information encoding of the data representing the color image, in which the CSF1 varies depending on luminance values associated with local regions of the data representing the color image, and in which the CSF1 is used for processing luminance data and the CSF2 is used for processing chrominance data; and
using one or more programmed processors, transforming the data representing the color image by encoding plural-bit information therein, in which the encoding is guided based on results obtained from the processing including predicted degradation of image areas.
20. The method of claim 19 in which the CSF1 varies spatially.
21. The method of claim 20 in which the CSF1 varies spatially in terms of spatial width.
22. The method of claim 20 in which the CSF2 varies spatially in terms of spatial width.
23. The method of claim 19 in which the processing of the data representing the color image with the CSF1 produces image blurring as the predicted degradation, in which the CSF1 varies so that relatively more blurring occurs as luminance of a local image region decreases.
24. The method of claim 19 in which the encoding is guided based on results obtained from the processing by varying embedding strength across different image areas of the data representing the color image based on predicted degradation of the different image areas.
25. The method of claim 24 in which predicted visibility of the encoding across the different image areas comprises uniform predicted visibility.
26. The method of claim 19 further comprising processing the data representing the color image with an attention model to predict visual traffic areas.
27. The method of claim 26 in which the encoding is guided based on the results obtained from the processing including predicted degradation of image areas and on predictions of visual traffic areas obtained from processing the obtained color image data with the attention model.
28. The method of claim 19 in which the encoding comprising digital watermarking.
29. The method of claim 19 in which the chrominance contrast sensitivity function (CSF2) comprises a blue-yellow contrast sensitivity function and a red-green contrast sensitivity function.
30. The method of claim 19 in which the CSF2 varies depending on luminance values associated with local regions of the data representing a color image.
31. An apparatus comprising:
an input to obtain color image data;
memory storing a luminance contrast sensitivity function (CSF1) and a chrominance contrast sensitivity function (CSF2); and
one or more processors configured for:
degrading data representing obtained color image data with the luminance contrast sensitivity function (CSF1) and the chrominance contrast sensitivity function (CSF2) to predict visibility changes attributable to encoding plural-bit information in the obtained color image data, in which the CSF1 varies depending on luminance values associated with local regions of the color image data, in which said degrading data representing the obtained color image data yields results for different image areas within the obtained color image data, and in which the CSF1 is used for degrading luminance data and the CSF2 is used for degrading chrominance data; and
transforming the obtained color image data by encoding plural-bit information therein, in which signal embedding strength of the encoding within the different image areas varies based on the results.
32. The apparatus of claim 31 in which the degrading of the data representing the obtained color image data with the CSF1 introduces image blurring, and in which the CSF1 varies so that relatively more blurring occurs as luminance of a local image region decreases.
33. The apparatus of claim 31 in which the varying embedding strength across different image regions of the color image data yields uniform predicted visibility of the encoding across the data representing the obtained color image data.
34. The apparatus of claim 31 in which the CSF1 varies spatially.
35. The apparatus of claim 34 in which the CSF1 varies spatially in terms of spatial width.
36. The apparatus of claim 31 in which the CSF2 varies spatially in terms of spatial width.
37. The apparatus of claim 31 in which the chrominance contrast sensitivity function (CSF2) comprises a blue-yellow contrast sensitivity function and a red-green contrast sensitivity function.
38. The apparatus of claim 31 in which the CSF2 varies depending on luminance values associated with local regions of the obtained color image data.