Lossless image compression using block based prediction and optimized context adaptive entropy coding

Patent metadata

Field	Value
Publication number	US-12010296-B2
Application number	US-202217891057-A
Country	US
Kind code	B2
Filing date	Aug 18, 2022
Priority date	Feb 17, 2021
Publication date	Jun 11, 2024
Grant date	Jun 11, 2024

Abstract

Embodiments are disclosed for lossless image compression using block-based prediction and context adaptive entropy coding. A method of lossless image compression using block-based prediction and context adaptive entropy coding comprises dividing an input image into a plurality of blocks, determining a pixel predictor for each block based on a block strategy, determining a plurality of residual values using the pixel predictor for each block, selecting a subset of features associated with the plurality of residual values, performing context modeling on the plurality of residual values based on the subset of features to identify a plurality of residual clusters, and entropy coding the plurality of residual clusters.

First claim

We claim:

1. A computer-implemented method comprising:
dividing an input image into a plurality of blocks, each block of the plurality of blocks being a portion of the input image;
determining a plurality of residual values using a pixel predictor for each block;
performing a classification of each feature of a set of features using a machine learning model feature selector trained to receive the input image and classify each feature of the set of features as belonging to a first subset of features or a second subset of features, wherein the set of features is defined by an image compression algorithm;
generating, by a machine learning context modeler, a residual cluster using the first subset of features and the plurality of residual values, wherein the generated residual cluster clusters one of more residual values of the plurality of residual values that have similar properties; and
entropy coding the residual clusters.

2. The method of claim 1 , wherein the plurality of blocks includes one or more blocks that are noncontiguous.

3. The method of claim 1 , wherein the pixel predictor determines a predicted pixel value by computing an average value of adjacent pixels by combining values of adjacent pixels.

4. The method of claim 1 , wherein the plurality of blocks includes one or more blocks that have different sizes.

5. The method of claim 1 , wherein generating, by the machine learning context modeler, the residual cluster using the first subset of features and the plurality of residual values, wherein the generated residual cluster clusters one of more residual values of the plurality of residual values that have similar properties:
identifying a set of properties using a lower resolution representation of the image; and
selecting from the set of properties, a third subset of features for input to the machine learning context modeler, the third subset of features representing a best compression result.

6. The method of claim 5 , wherein determining the plurality of residual values using a pixel predictor for each block comprises:
assigning a pixel predictor to a block in each color plane;
determining a compression performance of each pixel predictor for each block; and
identifying a subset of blocks for reassignment to a different pixel predictor using a threshold compression performance and the compression performance.

7. The method of claim 6 , wherein determining the compression performance comprises determining the compression performance on a representative version of the input image using a third subset of different sets of the features.

8. A system, comprising:
a memory component; and
a processing device coupled to the memory component, the processing device to perform operations comprising:
dividing an input image into a plurality of blocks, each block of the plurality of blocks being a portion of the input image;
determining a plurality of residual values using a pixel predictor for each block;
performing a classification of each feature of a set of features using a machine learning model feature selector trained to receive the input image and classify each feature of the set of features as belonging to a first subset of features or a second subset of features, wherein the set of features is defined by an image compression algorithm;
generating, by a machine learning context modeler, a residual cluster using the first subset of features and the plurality of residual values, wherein the generated residual cluster clusters one of more residual values of the plurality of residual values that have similar properties; and
entropy coding the residual clusters.

9. The system of claim 8 , wherein the plurality of blocks includes one or more blocks that are noncontiguous.

10. The system of claim 8 , wherein the pixel predictor determines a predicted pixel value by computing an average value of adjacent pixels by combining values of adjacent pixels.

11. The system of claim 8 , wherein the plurality of blocks includes one or more blocks that have different sizes.

12. The system of claim 8 , the operation for generating, by the machine learning context modeler, the residual cluster using the first subset of features and the plurality of residual values, wherein the generated residual cluster clusters one of more residual values of the plurality of residual values that have similar properties comprises operations further causing the processing device to perform operations comprising:
identifying a set of properties using a lower resolution representation of the image; and
selecting from the set of properties, a third subset of features for input to the machine learning context modeler, the third subset of features representing a best compression result.

13. The system of claim 12 , the operation for determining the plurality of residual values using a pixel predictor for each block comprises operations further causing the processing device to perform operations comprising:
assigning a pixel predictor to a block in each color plane;
determining a compression performance of each pixel predictor for each block; and
identifying a subset of blocks for reassignment to a different pixel predictor using a threshold compression performance and the compression performance.

14. The system of claim 13 , the operation determining the compression performance comprises operations further causing the processing device to perform operations comprising determining the compression performance on a representative version of the input image using a third subset of features.

15. A non-transitory computer-readable medium storing executable instructions, which when executed by a processing device, causes the processing device to perform operations comprising:
dividing an input image into a plurality of blocks, each block of the plurality of blocks being a portion of the input image;
determining a plurality of residual values using a pixel predictor for each block;
performing a classification of each feature of a set of features using a machine learning model feature selector trained to receive the input image and classify each feature of the set of features as belonging to a first subset of features or a second subset of features, wherein the set of features is defined by an image compression algorithm;
generating, by a machine learning context modeler, a residual cluster using the first subset of features and the plurality of residual values, wherein the generated residual cluster clusters one of more residual values of the plurality of residual values that have similar properties; and
entropy coding the residual clusters.

16. The non-transitory computer-readable medium of claim 15 , wherein the plurality of blocks includes one or more blocks that are noncontiguous.

17. The non-transitory computer-readable medium of claim 15 , wherein the pixel predictor determines a predicted pixel value by computing an average value of adjacent pixels by combining values of adjacent pixels.

18. The non-transitory computer-readable medium of claim 15 , wherein the plurality of blocks includes one or more blocks that have different sizes.

19. The non-transitory computer-readable medium of claim 15 , the instructions for generating, by the machine learning context modeler, the residual cluster using the first subset of features and the plurality of residual values, wherein the generated residual cluster clusters one of more residual values of the plurality of residual values that have similar properties further cause the processing device to per

Assignees

• Adobe Inc

Inventors

• Petrangeli Stefano
• Swaminathan Viswanathan
• Wang Haoliang

Classifications

• H04N19/176

  the region being a block, e.g. a macroblock · CPC title

• H04N19/91

  Entropy coding, e.g. variable length coding [VLC] or arithmetic coding · CPC title

• H04N19/182

  the unit being a pixel · CPC title

• H04N19/186

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

• H04N19/105 Primary

  Selection of the reference unit for prediction within a chosen coding or prediction mode, e.g. adaptive choice of position and number of pixels used for prediction · CPC title