Decoding with embedded denoising

• US 8,831,111 B2
• Filed: 05/16/2008
• Issued: 09/09/2014
• Est. Priority Date: 05/19/2006
• Status: Active Grant

• Current Assignee: Hong Kong University of Science and Technology
• Original Assignee: Hong Kong University of Science and Technology
• Inventors: Au, Oscar Chi Lim, Guo, Liwei, Wong, Hon Wah
• Primary Examiner(s): Sheikh, Ayaz
• Assistant Examiner(s): Ghafoerkhan, Faiyazkhan
• Application Number:
  US12/122,163
• Publication Number:
  US 20080285655A1
• Time in Patent Office: 2,307 Days
• Field of Search: None
• US Class Current: 375/240.29
• CPC Class Codes: H04N 19/44 Decoders specially adapted therefor, e.g. video decoders which are asymmetric with respect to the encoder
• CPC Class Codes: H04N 19/51 Motion estimation or motion compensation
• CPC Class Codes: H04N 19/577 Motion compensation with bidirectional frame interpolation, i.e. using B-pictures
• CPC Class Codes: H04N 19/80 Details of filtering operations specially adapted for video compression, e.g. for pixel interpolation H04N19/635, H04N19/86 take precedence

Abstract

Methods and systems for denoising embedded digital video decoding. Prediction and residue block of a current frame are obtained from motion vector. Variance of residue block is calculated using prior reference blocks, and a causal temporal linear minimum square error estimator is used to calculate a filter coefficient. The residue block is modified using the filter coefficient, and an output digital bitstream of blocks of pixels of the current frame is produced using the modified residue block and prior denoised prediction value of prior frames.

Claims

1. A method for decoding digital signals, comprising:

receiving an encoded digital signal;

estimating a signal statistic of the encoded digital signal; and

decoding the encoded digital signal by a decoder including a filter based at least in part on the signal statistic and a difference between a first version of a reference frame that is not denoised and a second version of the reference frame that is denoised by the filter included in the decoder.

2. The method of claim 1, wherein the estimating the signal statistic includes averaging noise variances of previously decoded reference frames.

3. The method of claim 2, wherein the decoding further includes:

determining a first prediction value for the encoded digital signal using the first version of the reference frame and a second prediction value for the digital signal using the second version of the reference frame;

calculating a correction coefficient using the noise variances and the signal statistic; and

modifying the first prediction value and the second prediction value using the correction coefficient to thereby produce a decoded digital signal.

4. The method of claim 3, wherein the determining the second prediction value includes averaging multiple blocks in the second version of the reference frame.

5. The method of claim 3, wherein the second version of the reference frame is received from a reference frame buffer.

6. The method of claim 3, wherein the calculating includes calculating the correction coefficient using a causal temporal linear minimum mean square error estimator (LMMSE).

7. The method of claim 6, wherein the modifying includes modifying the encoded digital signal based on a weighted linear combination of the first prediction value, the second prediction value and the correction coefficient.

8. The method of claim 1, further comprising receiving the first version of the reference frame from a first reference frame buffer in a first motion compensation module and receiving the second version of the reference frame from a second reference frame buffer in a second motion compensation module.

9. The method of claim 1, further comprising modifying a residue block received from the encoded digital signal based at least in part on the signal statistic.

10. The method of claim 9, wherein the modifying includes modifying the residue block based at least in part on the difference between the first version of the reference frame and the second version of the reference frame.

11. A method for decoding digital video signals, comprising

receiving an encoded bitstream that includes a video pixel;

calculating a signal statistic and a noise statistic of the video pixel; and

decoding the video pixel by a decoder associated with denoising filtering based at least in part on the signal statistic, the noise statistic and a difference between a first version of a reference frame that is not denoised and a second version of the reference frame that is denoised by the decoder associated with the denoising filtering.

12. The method of claim 11, wherein the calculating includes averaging noise variances of previously decoded pixels in the first version of the reference frame.

13. The method of claim 12, wherein the decoding includes:

determining a first prediction value of the video pixel using the first version of the reference frame and a second prediction value of the video pixel using the second version of the reference frame;

calculating a correction coefficient using the noise variances and the signal statistic; and

modifying the first prediction value and the second prediction value using the correction coefficient, thereby producing a decoded video pixel.

14. The method of claim 13, wherein the determining the second prediction value includes averaging blocks of pixels from different reference frames.

15. The method of claim 13, wherein the second version of the reference frame includes a previously denoised pixel or a block of previously denoised pixels.

16. The method of claim 13, wherein the first version of the reference frame includes denoised pixels and non-denoised pixels.

17. The method of claim 13, wherein the calculating the correction coefficient includes calculating the correction coefficient using a causal temporal linear minimum mean square error estimator (LMMSE).

18. The method of claim 17, wherein the modifying includes linearly combining the first prediction value, the second prediction value and the correction coefficient.

19. A method for decoding digital signals, comprising:

receiving an encoded digital signal;

estimating a first statistic associated with noise of the encoded digital signal and a second statistic associated with the first statistic; and

decoding the encoded digital signal by a decoder including a denoising filter based at least in part on the first statistic, the second statistic and a difference between a first version of a reference block that is not denoised and a second version of the reference block that is denoised by the denoising filter included in the decoder.

20. The method of claim 19, wherein the second statistic is generated based at least in part on the first statistic.