Architecture for signal enhancement coding

Patent metadata

Field	Value
Publication number	US-12212781-B2
Application number	US-201917265146-A
Country	US
Kind code	B2
Filing date	Aug 1, 2019
Priority date	Aug 3, 2018
Publication date	Jan 28, 2025
Grant date	Jan 28, 2025

Abstract

There is disclosed a method of encoding an input signal, the method comprising producing a base encoded signal by feeding an encoder with a down-sampled version of an input signal. The method further comprising producing a first quantised residual signal by: decoding the base encoded signal to produce a base decoded signal; and using a difference between the base decoded signal and the down-sampled version of the input signal to produce a first residual signal; quantising the first residual signal to produce the first quantised residual signal. The method further comprises producing a second residual signal by: de-quantising the first quantised residual signal to produce a reconstructed version of the first residual signal; correcting the base decoded signal using the first reconstructed version of the residual signal to create a corrected decoded version; upsampling the corrected decoded version; and using a difference between the corrected decoded signal and the input signal to produce the second residual signal.

First claim

The invention claimed is:

1. A method of encoding an input signal, the method comprising:
producing a base encoded signal by feeding an encoder with a down-sampled version of an input signal;
producing a first quantised residual signal by:
decoding the base encoded signal to produce a base decoded signal;
using a difference between the base decoded signal and the down-sampled version of the input signal to produce a first residual signal, wherein the first residual signal is subjected to a transform that involves the encoder selecting between use of a N×N kernel and a M×M kernel for application against the first residual signal, such that the encoder, as a part of said selection, considers both the N×N kernel and the M×M kernel, and wherein distributions of both the first residual signal and a second residual signal have mass ranges that are within a threshold around 0, such that level 0 and level 1 enhancement streams have their masses at about a value of 0; and
quantising the first residual signal to produce the first quantised residual signal, wherein said quantising includes converting a division operation that involves a selected step-width into a multiplication operation that involves an inverse of the step-width and then applying the multiplication operation to the first residual signal;

producing the second residual signal by:
de-quantising the first quantised residual signal to produce a reconstructed version of the first residual signal;
correcting the base decoded signal using the first reconstructed version of the residual signal to create a corrected decoded version;
upsampling the corrected decoded version; and
using a difference between the corrected decoded signal and the input signal to produce the second residual signal,

wherein the input signal when encoded comprises the base encoded signal, the first quantised signal and the second residual signal.

2. The method of claim 1 , wherein the step of producing a first quantised residual signal comprises performing said transform to the first residual signal prior to quantising to produce a first transformed and quantised residual signal, and the step of producing a second residual signal comprises inverse transforming the first transformed and quantised signal after de-quantising the first transformed and quantised residual signal.

3. The method of claim 2 , wherein said transform is a directional decomposition transform that is applied to the first residual signal to create components that correspond to the first residual signal.

4. The method of claim 3 , where the directional decomposition transform comprises the M×M kernel, which is a 4×4 kernel.

5. The method of claim 3 , wherein the directional decomposition transform is a Hadamard-based transform.

6. The method of claim 3 , wherein the transform produces the following components: average, vertical, horizontal and diagonal.

7. The method of claim 1 , wherein the step of producing the second residual signal comprises quantising the second residual signal to produce a second quantised residual signal.

8. The method of claim 7 , wherein the step of producing the second residual signal comprises transforming the second residual signal prior to quantising the second residual signal to produce a second transformed and quantised signal.

9. The method of claim 8 , wherein the step of transforming the second residual signal comprises applying a directional decomposition transform to the second residual signal to create components that correspond to the second residual signal.

10. The method of claim 9 , where the directional decomposition transform comprises the M×M kernel, which is a 4×4 kernel.

11. The method of claim 9 , wherein the directional decomposition transform is a Hadamard-based transform.

12. The method of claim 9 , wherein the transform produces the following components: average, vertical, horizontal and diagonal.

13. The method of claim 1 , wherein one or more of the first quantised residual signal and the second residual signal is entropy encoded.

14. The method of claim 13 , wherein the entropy encoding comprises one or both of run-length encoding and Huffman encoding.

15. The method of claim 1 , wherein the input signal is a video signal.

16. The method of claim 15 , wherein the input signal is a frame of a video signal.

17. A non-transitory computer-readable storage medium comprising instructions which when executed by a processor cause the processor to:
produce a base encoded signal by feeding an encoder with a down-sampled version of an input signal;
produce a first quantised residual signal by:
decoding the base encoded signal to produce a base decoded signal;
using a difference between the base decoded signal and the down-sampled version of the input signal to produce a first residual signal, wherein the first residual signal is subjected to a transform, and wherein the transform involves the encoder selecting between use of a 2×2 kernel and a 4×4 kernel for application against the first residual signal, such that the encoder, as a part of said selection, considers both the 2×2 kernel and the 4×4 kernel, and wherein distributions of both the first residual signal and a second residual signal have mass ranges that are within a threshold around 0, such that level 0 and level 1 enhancement streams have their masses at about a value of 0; and
quantising the first residual signal to produce the first quantised residual signal, wherein said quantising includes converting a division operation that involves a selected step-width into a multiplication operation that involves an inverse of the step-width and then applying the multiplication operation to the first residual signal;

produce the second residual signal by:
de-quantising the first quantised residual signal to produce a reconstructed version of the first residual signal;
correcting the base decoded signal using the first reconstructed version of the residual signal to create a corrected decoded version;
upsampling the corrected decoded version; and
using a difference between the corrected decoded signal and the input signal to produce the second residual signal,
wherein the second residual signal is structured to correct artifacts introduced from a use of the first residual signal, and
wherein the artifacts corrected by the second residual signal are artifacts that are generated over a spatial scale having an area that is equal to or greater than 4 pixels;

wherein the input signal when encoded comprises the base encoded signal, the first quantised signal and the second residual signal.

18. An encoder comprising a processor and memory, the processor configured to:
produce a base encoded signal by feeding an encoder with a down-sampled version of an input signal;
produce a first quantised residual signal by:
decoding the base encoded signal to produce a base decoded signal;
using a difference between the base decoded signal and the down-sampled version of the input signal to produce a first residual signal, wherein the first residual signal is subjected to a transform that involves the encoder selecting between use of a 2×2 kernel and a 4×4 kernel for application against the first residual signal, such that the encoder, as a part of said selection, considers both the 2×2 kernel and the 4×4 kernel, and wherein distributions of both the first residual signal and a second residual signal have mass ranges that are within a threshold around 0, such that level 0 and level 1 enhancement streams have their masses at about a value of 0; and
quantising t

Assignees

• V Nova Int Ltd

Inventors

• Meardi Guido
• Ciccarelli Lorenzo
• Ferrara Simone

Classifications

• H04N19/91

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

• H04N19/60

  using transform coding · CPC title

• H04N19/93

  Run-length coding · CPC title

• H04N19/70

  characterised by syntax aspects related to video coding, e.g. related to compression standards · CPC title

• H04N19/122

  Selection of transform size, e.g. 8x8 or 2x4x8 DCT; Selection of sub-band transforms of varying structure or type · CPC title