High precision encoding and decoding of video images

Patent metadata

Field	Value
Publication number	US-10264256-B2
Application number	US-201314101759-A
Country	US
Kind code	B2
Filing date	Dec 10, 2013
Priority date	Mar 2, 2001
Publication date	Apr 16, 2019
Grant date	Apr 16, 2019

Abstract

Methods, systems, and computer programs for improved quality video compression. Image quality from MPEG-style video coding may be improved by preserving a higher number of bits during intermediate encoding and decoding processing steps. Problems of inverse discrete cosine transform (IDCT) mismatch can be eliminated by exactly matching the IDCT function numerical algorithm of the decoder to the IDCT function numerical algorithm used for the decoding portion of the encoder. Also included is an application of high precision compression to wide dynamic range images by extending the range of the “quantization parameter” or “QP”. The extension of QP may be accomplished either by increasing the range of QP directly, or indirectly through a non-linear transformation. Also included is an application of extended intermediate processing precision and an extended QP range to reduced contrast regions of an image to extend the precision with which the low-contrast portions are compression coded.

First claim

What is claimed is:

1. A non-transitory computer-readable medium storing software comprising instructions executable by one or more computers which, upon such execution, cause the one or more computers to perform operations comprising:
obtaining, by a decoder and from a bitstream of a compressed picture, a representational quantization parameter code for a pixel region of the compressed picture, the representational quantization parameter code selected from a first set of levels of representational quantization parameter codes, the first set of levels of representational quantization parameter codes having at least 32 level values, wherein a count of the level values in the first set of levels of representational quantization parameter codes depends on a bit precision of the compressed picture;
mapping, by the decoder and using a non-linear transformation, the obtained representational quantization parameter code from the first set of levels to a second set of levels for determined quantization parameter values, the second set of levels being different than the first set of levels, each determined quantization parameter value defining a respective transformation to be applied to quantized transform coefficients during decoding of the compressed picture; and
decoding, by the decoder, the pixel region using a determined quantization parameter value that maps to the obtained representational quantization parameter code, comprising applying the transformation defined by the determined quantization parameter value to quantized transform coefficients associated with the pixel region.

2. The computer-readable medium of claim 1 , wherein the non-linear transformation is linear for mapping a first portion of the first set of levels and non-linear in mapping a second portion of the first set of levels, the second portion comprising level values that are greater than the level values of the first portion.

3. The computer-readable medium of claim 1 , wherein the second set of levels comprises at least 128 level values.

4. The computer-readable medium of claim 1 , wherein the compressed picture is a wide dynamic range picture that has a pixel bit precision greater than 8 bits.

5. The computer-readable medium of claim 1 , wherein the operations further comprise obtaining the non-linear transformation using a lookup table.

6. The computer-readable medium of claim 1 , wherein the non-linear transformation comprises a logarithmic-type function.

7. A system for decoding a picture comprising:
one or more computers and one or more storage devices storing instructions that are operable, when executed by the one or more computers, to cause the one or more computers to perform operations comprising:
obtaining, by a decoder and from a bitstream of a coded picture, a representational quantization parameter code for a pixel region of a compressed picture, the representational quantization parameter code selected from a first set of levels of representational quantization parameter codes, the first set of levels of representational quantization parameter codes having at least 32 level values, wherein a count of the level values in the first set of levels of representational quantization parameter codes depends on a bit precision of the picture;
mapping, by the decoder and using a non-linear transformation, the obtained representational quantization parameter code from the first set of levels to a second set of levels for determined quantization parameter values, the second set of levels being different than the first set of levels, each determined quantization parameter value defining a respective transformation to be applied to quantized transform coefficients during decoding of the compressed picture; and
decoding, by the decoder, the pixel region using a determined quantization parameter value that maps to the obtained representational quantization parameter code, comprising applying the transformation defined by the determined quantization parameter value to quantized transform coefficients associated with the pixel region.

8. The system of claim 7 , wherein the non-linear transformation is linear for mapping a first portion of the first set of levels and non-linear in mapping a second portion of the first set of levels, the second portion comprising level values that are greater than the level values of the first portion.

9. The system of claim 7 , wherein the second set of levels comprises at least 128 level values.

10. The system of claim 7 , wherein the compressed picture is a wide dynamic range picture that has a pixel bit precision greater than 8 bits.

11. The system of claim 7 , wherein the operations further comprise obtaining the non-linear transformation using a lookup table.

12. The system of claim 7 , wherein the non-linear transformation comprises a logarithmic-type function.

13. A computer-implemented method of decoding a picture comprising:
obtaining, by a decoder and from a bitstream of a coded picture, a representational quantization parameter code for a pixel region of a compressed picture, the representational quantization parameter code selected from a first set of levels of representational quantization parameter codes, the first set of levels of representational quantization parameter codes having at least 32 level values, wherein a count of the level values in the first set of levels of representational quantization parameter codes depends on a bit precision of the picture;
mapping, by the decoder and using a non-linear transformation, the obtained representational quantization parameter code from the first set of levels to a second set of levels for determined quantization parameter values, the second set of levels being different than the first set of levels, each determined quantization parameter value defining a respective transformation to be applied to quantized transform coefficients during decoding of the compressed picture; and
decoding, by the decoder, the pixel region using a determined quantization parameter value that maps to the obtained representational quantization parameter code, comprising applying the transformation defined by the determined quantization parameter value to quantized transform coefficients associated with the pixel region.

14. The method of claim 13 , wherein the non-linear transformation is linear for mapping a first portion of the first set of levels and non-linear in mapping a second portion of the first set of levels, the second portion comprising level values that are greater than the level values of the first portion.

15. The method of claim 13 , wherein the second set of levels comprises at least 128 level values.

16. The method of claim 13 , wherein the compressed picture is a wide dynamic range picture that has a pixel bit precision greater than 8 bits.

17. The method of claim 13 , further comprising obtaining the non-linear transformation using a lookup table.

18. The method of claim 13 , wherein the non-linear transformation comprises a logarithmic-type function.

Assignees

• Dolby Laboratories Licensing Corp

Inventors

• Demos Gary A
• Ruhoff David

Classifications

• H04N19/60

  using transform coding · CPC title

• H04N19/90

  using coding techniques not provided for in groups H04N19/10-H04N19/85, e.g. fractals · CPC title

• H04N19/182

  the unit being a pixel · CPC title

• H04N19/136

  Incoming video signal characteristics or properties · CPC title

• H04N19/196

  being specially adapted for the computation of encoding parameters, e.g. by averaging previously computed encoding parameters (processing of motion vectors H04N19/513) · CPC title