PREDICTIVE QUANTIZATION CODING METHOD AND VIDEO COMPRESSION SYSTEM

• US 20200137392A1
• Filed: 12/28/2018
• Published: 04/30/2020
• Est. Priority Date: 10/26/2018
• Status: Active Grant

• Current Assignee: IP3 2017, Series 200 of Allied Security Trust I (Allied Security Trust)
• Original Assignee: Xi'An Creation Keji Co., Ltd.
• Inventors: YUE, QINGDONG, RAN, WENFANG, LI, WEN
• Granted Patent: US 10,645,387 B1
• Time in Patent Office: Days
• CPC Class Codes: H04N 19/105 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 Class Codes: H04N 19/11 among a plurality of spatial predictive coding modes
• CPC Class Codes: H04N 19/119 Adaptive subdivision aspects, e.g. subdivision of a picture into rectangular or non-rectangular coding blocks
• CPC Class Codes: H04N 19/126 Details of normalisation or weighting functions, e.g. normalisation matrices or variable uniform quantisers
• CPC Class Codes: H04N 19/147 according to rate distortion criteria rate-distortion as a criterion for motion estimation H04N19/567
• CPC Class Codes: H04N 19/182 the unit being a pixel
• CPC Class Codes: H04N 19/186 the unit being a colour or a chrominance component
• CPC Class Codes: H04N 19/50 using predictive coding H04N19/61 takes precedence

Abstract

The present invention relates to a predictive quantization coding method and a video compression system. The method includes: dividing a pixel to be processed into a plurality of pixel components; obtaining one pixel component to be processed and texture direction gradients thereof; obtaining reference pixels and a prediction residual of the pixel component to be processed; forming a prediction residual code stream; dividing the prediction residual code stream into multiple quantization units; and obtaining a quantization residual code stream. The present invention can reduce the transmission bandwidth, and reduce the theoretical limit entropy and complexity.

Claims

1. A predictive quantization coding method, comprising steps of:

(a) dividing a pixel to be processed into a plurality of pixel components, wherein pixels of an image are sequentially taken as the pixel to be processed;

(b) obtaining one pixel component to be processed from the plurality of pixel components;

(c) obtaining texture direction gradients of the pixel component to be processed;

(d) obtaining reference pixels according to the texture direction gradients and positional relationships between the pixel component to be processed and a remaining of the plurality of pixel components;

(e) obtaining a prediction residual of the pixel component to be processed according to the reference pixels;

(f) repeating steps (b) to (e), and taking each pixel component of the plurality of pixel components and obtaining the prediction residual corresponding thereto, and forming a prediction residual code stream including the prediction residuals of the pixel to be processed of the image;

(g) dividing the prediction residual code stream into a plurality of quantization units each including a predetermined number of prediction residuals divided from the prediction residual code stream; and

(h) obtaining first rate distortion optimizations and second rate distortion optimizations corresponding to the plurality of quantization units to obtain a quantization residual code stream.

2. The predictive quantization coding method according to claim 1, wherein the step (a) of dividing the pixel to be processed into a plurality of pixel components comprises:

dividing the pixel to be processed into a R pixel component, a G pixel component, and a B pixel component.

3. The predictive quantization coding method according to claim 1, wherein the step (d) comprises following sub-steps of:

(d1) obtaining a first weighting gradient optimal value according to the texture direction gradients;

(d2) obtaining a second weighting gradient optimal value according to the first weighting gradient optimal value and the positional relationships between the pixel component to be processed and the remaining of the plurality of pixel components; and

(d3) obtaining the reference pixels according to the second weighting gradient optimal value.

4. The predictive quantization coding method according to claim 1, wherein the sub-step (d2) comprises:

(d21) obtaining positional relationship weights according to the positional relationships between the pixel component to be processed and the remaining of the plurality of pixel components; and

(d22) obtaining the second weighting gradient optimal value according to the positional relationship weights and the first weighting gradient optimal value.

5. The predictive quantization coding method according to claim 1, wherein the positional relationships between the pixel components to be processed and the remaining of the plurality of pixel components satisfy that:

the closer the pixel component to the pixel component to be processed is, the greater the positional relationship weight is, otherwise is smaller.

6. The predictive quantization coding method according to claim 1, wherein the step (h) comprises sub-steps of:

(h1) performing quantization processing on a prediction residual of each of the quantization units to obtain a quantization residual;

(h2) sequentially performing a first inverse quantization processing and a first compensation processing on the quantization residual, to obtain a first inverse quantization residual and a first rate distortion optimization;

(h3) performing a second compensation processing on the first inverse quantization residual, to obtain a second inverse quantization residual and a second rate distortion optimization;

(h4) comparing the first rate distortion optimization and the second rate distortion optimization, setting a compensation flag bit to be no compensation if the first rate distortion optimization is less than the second rate distortion optimization;

otherwise, setting the compensation flag bit to be compensation; and

(h5) writing the compensation flag bit and the quantization residual into the quantization residual code stream.

7. The predictive quantization coding method according to claim 6, wherein the sub-step (h2) comprises:

(h21) sequentially performing the first inverse quantization processing and the first compensation processing on the quantization residual, to obtain the first inverse quantization residual; and

(h22) obtaining the first rate distortion optimization according to the first inverse quantization residual, the prediction residuals, and the quantization residual.

8. The predictive quantization coding method according to claim 7, wherein the sub-step (h3) comprises:

(h31) obtaining a fluctuation coefficient according to a first residual loss;

(h32) performing the second compensation processing on the first inverse quantization residual according to the fluctuation coefficient and a fluctuation state, to obtain the second inverse quantization residual; and

(h33) obtaining the second rate distortion optimization according to the second inverse quantization residual, the prediction residual, and the quantization residual.

9. The predictive quantization coding method according to claim 8, wherein the fluctuation coefficient k satisfies:

10. A video compression system, comprising:

a memory and at least one processor coupled to the memory, wherein the at least one processor is configured to perform the predictive quantization coding method according to claim 1.