Methods and apparatus for adaptive coupled pre-processing and post-processing filters for video encoding and decoding

Patent metadata

Field	Value
Publication number	US-9883207-B2
Application number	US-201013518942-A
Country	US
Kind code	B2
Filing date	Dec 2, 2010
Priority date	Dec 31, 2009
Publication date	Jan 30, 2018
Grant date	Jan 30, 2018

Abstract

Methods and apparatus are provided for adaptive coupled pre-processing and post-processing filters for video encoding and decoding. The apparatus and method encode input data for a picture into a resultant bitstream, wherein said video encoder comprises a pre-filter and a post-filter coupled to the pre-filter, wherein said pre-filter filters the input data for the picture and the post-filter filters in-loop reconstructed data for the picture.

First claim

The invention claimed is:

1. An apparatus, comprising:
a video encoder for encoding input data for a picture into a resultant bitstream, wherein said video encoder comprises a pre-filter and a post-filter coupled to the pre-filter, wherein said pre-filter filters the input data for the picture and the post-filter filters in-loop reconstructed data for the picture,
wherein at least one of filter coefficients and filter parameters for at least one of the pre-filter and the post-filter are selected responsive to at least one of a resolution, a local gradient, and a gradient direction, wherein gradient is a difference between pixels being pre-filtered, and adaptation is made discrete by partitioning a gradient strength, gradient direction and gradient variance into disjoint ranges and applying fixed filters for different ranges, and
wherein a Lagrangian parameter for luminance and chrominance in the video encoder's rate-distortion optimizer is a function of the coding efficiency of the pre-filter and the post-filter; and
wherein the pre-filter and post-filter are implemented with a lifting scheme and adaptation is embedded therein by changing values of a diagonal matrix, obtained by performing singular value decomposition on a fixed pre-filter transform matrix, such that values of the diagonal matrix can be reversed at a decoder.

2. In a video encoder, a method, comprising:
encoding input data for a picture into a resultant bitstream, wherein said video encoder comprises a pre-filter and a post-filter coupled to the pre-filter, wherein said pre-filter filters the input data for the picture and the post-filter filters in-loop reconstructed data for the picture,
wherein at least one of filter coefficients and filter parameters for at least one of the pre-filter and the post-filter are selected responsive to at least one of a resolution, a local gradient, and a gradient direction, wherein gradient is a difference between pixels being pre-filtered, and adaptation is made discrete by partitioning a gradient strength, gradient direction and gradient variance into disjoint ranges and applying fixed filters for different ranges, and
wherein a Lagrangian parameter for luminance and chrominance in the video encoder's rate-distortion optimizer is a function of the coding efficiency of the pre-filter and the post-filter; and
wherein the pre-filter and post-filter are implemented with a lifting scheme and adaptation is embedded therein by changing values of a diagonal matrix, obtained by performing singular value decomposition on a fixed pre-filter transform matrix, such that values of the diagonal matrix can be reversed at a decoder.

3. The method of claim 2 , wherein filter coefficients and filter parameters of the post-filter are selected such that the post-filter is an exact inverse of the pre-filter.

4. The method of claim 2 , wherein at least one of the pre-filter and the post-filter comprise different filters, and chroma components of the picture are filtered using different ones of the different filters than the luma components of the picture.

5. The method of claim 2 , wherein a post-filter transform matrix used by the post-filter to filter the reconstructed data is decomposed into two summands, one of the two summands being an identity matrix and another one of the two summands being a matrix representing a multiplication of a first, a second, and a third matrix, the first and the third matrices being fixed, and the second matrix being variable in order to reverse an adaptation of the pre-filter.

6. The method of claim 2 , wherein filter coefficients and filter parameters of the post-filter are selected such that the post-filter is an inverse of the pre-filter.

7. The method of claim 6 , wherein the post-filter is configured to provide a same output data there from as input data provided to the pre-filter by minimizing a difference between an observation and a pre-filtered estimate, the observation relating to input data provided to the post-filter, and the pre-filtered estimate relating to an estimate of the input data for the picture prior to filtering by the pre-filter.

8. The method of claim 2 , wherein the pre-filter and the post-filter are integer implementations determined so as to minimize a distance relating to an exact invertibility between the pre-filter and the post-filter.

9. The method of claim 2 , wherein a filter size of the pre-filter and the post-filter is a same size as a transform applied to residue data, the residue data representing a difference between the input data for the picture and reference data for at least one reference picture.

10. The method of claim 9 , wherein the pre-filter and the post-filer comprise multiple filters, and at least one of the multiple filters is applied to all transform sizes of transforms applied to the residue data.

11. The method of claim 2 , wherein a filter size of at least one of the pre-filter and the post-filter is different from a size of a transform applied to the input data.

12. The method of claim 2 , wherein the pre-filter and the post-filter are applied to only a portion of the input data.

13. The method of claim 12 , wherein the portion of the input data is selected from at least one of a block boundary, and within a block.

14. An apparatus, comprising:
a video decoder for decoding residual image data for a picture, wherein said video decoder comprises a pre-filter and a post-filter coupled to the pre-filter, wherein said pre-filter filters a reference picture for use in decoding the residual image data and the post-filter filters in-loop reconstructed data for the picture,
wherein at least one of filter coefficients and filter parameters for at least one of the pre-filter and the post-filter are selected responsive to at least one of a resolution, a local gradient, and a gradient direction, wherein gradient is a difference between pixels being pre-filtered, and adaptation is made discrete by partitioning a gradient strength, gradient direction and gradient variance into disjoint ranges and applying fixed filters for different ranges, and
wherein a Lagrangian parameter for luminance and chrominance in the video decoder's rate-distortion optimizer is a function of the coding efficiency of the pre-filter and the post-filter; and
wherein the pre-filter and post-filter are implemented with a lifting scheme and adaptation is embedded therein by changing values of a diagonal matrix, obtained by performing singular value decomposition on a fixed pre-filter transform matrix, such that values of the diagonal matrix can be reversed from those of an encoder.

15. In a video decoder, a method comprising:
decoding residual image data for a picture, wherein said video decoder comprises a pre-filter and a post-filter coupled to the pre-filter, wherein said pre-filter filters a reference picture for use in decoding the residual image data and the post-filter filters in-loop reconstructed data for the picture,
wherein at least one of filter coefficients and filter parameters for at least one of the pre-filter and the post-filter are selected responsive to at least one of a resolution, a local gradient, and a gradient direction, wherein gradient is a difference between pixels being pre-filtered, and adaptation is made discrete by partitioning a gradient strength, gradient direction and gradient variance into disjoint ranges and applying fixed filters for different ranges, and
wherein a Lagrangian parameter for luminance and chrominance in the video decoder's rate-distortion optimizer is a function of the coding efficiency of the pre-filter and the post-filter; and
wherein th

Assignees

• Misra Kiran
• Sole Joel
• Yin Peng
• Lu Xiaoan
• Zheng Yunfei
• Xu Qian
• Thomson Licensing Dtv

Inventors

• Misra Kiran
• Sole Joel
• Yin Peng
• Lu Xiaoan
• Zheng Yunfei
• Xu Qian

Classifications

• H04N19/82

  involving filtering within a prediction loop · CPC title

• H04N19/102

  characterised by the element, parameter or selection affected or controlled by the adaptive coding · CPC title

• H04N19/44

  Decoders specially adapted therefor, e.g. video decoders which are asymmetric with respect to the encoder · CPC title

• H04N19/85

  using pre-processing or post-processing specially adapted for video compression · CPC title

• H04N19/186

  the unit being a colour or a chrominance component · CPC title