Adaptive updates in motion-compensated temporal filtering

• US 8,442,108 B2
• Filed: 05/10/2005
• Issued: 05/14/2013
• Est. Priority Date: 07/12/2004
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Song, Li, Xu, Jizheng, Wu, Feng
• Primary Examiner(s): SENFI, BEHROOZ M
• Application Number:
  US11/126,997
• Publication Number:
  US 20060008038A1
• Time in Patent Office: 2,926 Days
• Field of Search: 348/189, 348/192, 348/180, 348/700, 348/699, 348/416, 348/412, 348/413, 348/398, 348/407, 348/403, 348/405, 348/415, 348/384, 348/390, 348/401, 348/402, 348/409, 382/232, 382/260, 382/254, 382/275, 382/274, 382/239, 382/240, 382/286, 375/240.12, 375/240.19, 375/240.24, 375/240.16, 375/240.08, 375/350
• US Class Current: 375/240.01
• CPC Class Codes: H04N 19/124 Quantisation
• CPC Class Codes: H04N 19/13 Adaptive entropy coding, e.g. adaptive variable length coding [AVLC] or context adaptive binary arithmetic coding [CABAC]
• CPC Class Codes: H04N 19/577 Motion compensation with bidirectional frame interpolation, i.e. using B-pictures
• CPC Class Codes: H04N 19/61 in combination with predictive coding
• CPC Class Codes: H04N 19/615 using motion compensated temporal filtering [MCTF]
• CPC Class Codes: H04N 19/63 using sub-band based transform, e.g. wavelets

Abstract

Techniques and tools are described for scalable video coding and decoding. For example, a video encoder and decoder perform adaptive update operations in lifting-based motion-compensated temporal filtering and inverse motion-compensated temporal filtering, respectively. The encoder, in a prediction stage, computes a motion-compensated prediction for an odd image and computes a high-pass image as the difference between the odd image and the prediction. For an update stage, the encoder performs motion compensation on the high-pass image (and potentially another high-pass image) to compute a motion-compensated update. Based upon perceptual criteria, the encoder adaptively combines the update with an even image to form a low-pass image. The decoder, for an update stage, performs motion compensation on one or more high-pass images to compute a motion-compensated update. Based upon perceptual criteria, the encoder adaptively combines the update with a low-pass image to reconstruct an even frame.

Claims

1. In a computing device that implements a video encoder, a method comprising:

with the computing device that implements the video encoder, performing motion-compensated temporal filtering that includes a motion-compensated prediction stage and a motion-compensated update stage, the motion-compensated update stage comprising;

adapting update information based at least in part upon a model of the human visual system and based at least in part upon sample values of an image to which the update information is to be applied, wherein the adapting includes setting one or more image-dependent thresholds from the sample values of the image to limit the update information for parts of the image in which distortion is especially perceptible and not limit the update information for other parts of the image; and

applying results of the adapting to the image during encoding.

2. The method of claim 1 wherein the model is a just noticeable difference model.

3. The method of claim 1 wherein the video encoder and a corresponding video decoder use the same metric for the adapting.

4. The method of claim 1 wherein the one or more image-dependent thresholds vary depending on variance of the sample values in a spatial neighborhood within the image.

5. The method of claim 1 wherein the one or more image-dependent thresholds are lower for flat areas and higher for noisy areas within the image.

6. The method of claim 1 wherein the one or more image-dependent thresholds vary depending on a contrast adjustment tuning parameter, and wherein the one or more image-dependent thresholds are further modified by a strength parameter.

7. The method of claim 1 wherein the adapting comprises:

passing through a part of the update information less than absolute value of an applicable threshold of the one or more image-dependent thresholds; and

otherwise, limiting the part of the update information to the applicable threshold as modified by sign of the part of the update information.

8. The method of claim 1 wherein the image to which the update information is to be applied is a second image, the method further comprising, during encoding:

in the motion-compensated prediction stage, for a first image,

computing a motion-compensated prediction for the first image with reference to at least the second image;

computing a high-pass image as the difference between the first image and the motion-compensated prediction for the first image; and

performing motion compensation on the high-pass image, wherein the motion-compensated high-pass image contributes to the update information, wherein the applying include combining the update information with the second image to form a low-pass image, and wherein the method further includes subsequent spatial decomposition and/or entropy encoding.

18. The method of claim 1 wherein the sample values of the image to which the update information is to be applied are spatial-domain values or low-pass sub-band values, the motion-compensated temporal filtering being spatial-domain motion-compensated temporal filtering or in-band motion-compensated temporal filtering.

9. A computer-readable storage medium storing computer-executable instructions for causing a computer system programmed thereby to perform a method comprising:

based at least in part on one or more perceptual criteria and based at least in part upon sample values of an image to which update information is to be applied in a motion-compensated update stage, computing one or more update thresholds for lifting-based motion-compensated temporal filtering or inverse motion-compensated temporal filtering, wherein the one or more update thresholds are image-dependent thresholds computed from the sample values of the image to limit the update information for parts of the image in which distortion is especially perceptible and not limit the update information for other parts of the image; and

thresholding update information for the motion-compensated update stage according to the one or more update thresholds.

10. The computer-readable storage medium of claim 9 wherein the one or more update thresholds vary depending on variance of the sample values in a spatial neighborhood within the image.

11. The computer-readable storage medium of claim 9 wherein the one or more update thresholds are lower for flat areas and higher for noisy areas within the image.

12. The computer-readable storage medium of claim 9 wherein the one or more update thresholds vary depending on a contrast adjustment tuning parameter.

13. The computer-readable storage medium of claim 9 wherein the one or more update thresholds vary depending on a strength parameter.

14. The computer-readable storage medium of claim 9 wherein the thresholding comprises:

passing through a part of the update information less than an applicable update threshold of the one or more update thresholds; and

otherwise, limiting the part of the update information to the applicable update threshold as modified by sign of the part of the update information.

15. A video encoder including a processor, memory and the computer-readable storage medium of claim 9.

16. A video decoder including a processor, memory and the computer-readable storage medium of claim 9.

19. The computer-readable storage medium of claim 9 wherein:

a high-pass image results from computing differences, in a motion-compensated prediction stage, between a first image and a motion-compensated prediction of the first image from at least a second image, the second image being the image to which the update information is to be applied;

the update information results from motion-compensated prediction from at least the high-pass image in the motion-compensated update stage; and

a low-pass image results from applying the thresholded update information to the second image.

20. The computer-readable storage medium of claim 9 wherein the sample values of the image to which the update information is to be applied are spatial-domain values or low-pass sub-band values.

17. A tool comprising:

one of more buffers for storing video picture information to which update information is to be applied; and

means for adaptively performing update operations in a motion-compensated update stage of motion-compensated temporal filtering or inverse motion-compensated temporal filtering in view of update thresholds computed based upon perceptual criteria and based upon sample values of the buffered video picture information to which the update information is to be applied, the update information being selectively applied according to the update thresholds, wherein the sample values considered during the motion-compensated update stage of motion-compensated temporal filtering are original values but the sample values considered during the motion-compensated update stage of inverse motion-compensated temporal filtering are reconstructed values.

21. In a computing device that implements a video decoder, a method comprising:

with the computing device that implements the video decoder, performing inverse motion-compensated temporal filtering that includes a motion-compensated prediction stage and a motion-compensated update stage, the motion-compensated update stage comprising;

adapting update information based at least in part upon a model of the human visual system and sample values of an image to which the update information is to be applied, wherein the adapting includes setting one or more image-dependent thresholds from the sample values of the image to limit the update information for parts of the image in which distortion is especially perceptible and not limit the update information for other parts of the image; and

applying results of the adapting to the image during decoding.

22. The method of claim 21 wherein the model is a just noticeable difference model.

23. The method of claim 21 wherein the one or more image-dependent thresholds vary depending on variance of the sample values in a spatial neighborhood within the image.

24. The method of claim 21 wherein the one or more image-dependent thresholds are lower for flat areas and higher for noisy areas within the image.

25. The method of claim 21 wherein the one or more image-dependent thresholds vary depending on a contrast adjustment tuning parameter, and wherein the one or more image-dependent thresholds are further modified by a strength parameter.

26. The method of claim 21 wherein the adapting comprises:

passing through a part of the update information less than absolute value of an applicable threshold of the one or more image-dependent thresholds; and

otherwise, limiting the part of the update information to the applicable threshold as modified by sign of the part of the update information.

27. The method of claim 21 further comprising, during decoding, performing motion compensation on one or more high-pass images, wherein the one or more motion-compensated high-pass images contribute to the update information, wherein the applying includes combining the update information with a low-pass image, and wherein the method further comprises subsequent motion compensated-prediction and/or display.

28. The method of claim 21 wherein the sample values of the image to which the update information is to be applied are spatial-domain values or low-pass sub-band values, the inverse motion-compensated temporal filtering being spatial-domain inverse motion-compensated temporal filtering or in-band inverse motion-compensated temporal filtering.