Adaptive Coding Group Size for Non-Square Transform Units

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The existing High Efficiency Video Coding (HEVC) standard faces inefficiencies in transform coefficient coding, particularly for non-square blocks, which affects storage and transmission bandwidth requirements for video data.

Innovation Solution

The method involves adaptively determining the size of coding groups (CGs) based on the transform unit (TU) size, allowing for non-square CG sizes such as 2x4 or 4x2 for non-square TUs, and optimizing the coefficient scan paths to improve coding efficiency and throughput.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If fixed-size coding groups (4x4) are used for all transform units, then the coding structure is simple and uniform, but the coding efficiency for non-square blocks is poor

Engineering Contradiction:
Improvecoding structure complexityVSAvoidcoding efficiency
Core Design Contradiction:
Device complexityVSProductivity

Solution Approach 1:

The patent applies dynamics by making the coding group size adaptive rather than fixed. The coding group dimensions are dynamically adjusted based on the transform unit aspect ratio, allowing the coding structure to adapt to different block shapes (square, horizontal, or vertical rectangles), thereby improving coding efficiency without significantly increasing system complexity

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The patent changes the parameter of coding group size from a fixed value (4x4) to variable values that depend on the transform unit dimensions. By calculating aspect ratios and selecting appropriate coding group sizes (e.g., 4x2 for vertical rectangles, 2x4 for horizontal rectangles), the system optimizes coding efficiency for different block shapes

Inventive Principle:
Principle #35Parameter changes

2Adaptability or versatility

If mode switching between context-coded and bypass-coded bins is frequently performed, then coding adaptability is improved, but processing overhead and complexity increase

Engineering Contradiction:
Improvecoding adaptabilityVSAvoidprocessing overhead
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent segments the transform unit into coding groups with sizes matched to the block shape, and further segments the coefficient coding into distinct scan paths. By organizing coefficients into vertical or horizontal scan paths based on the transform unit aspect ratio, the patent reduces the need for frequent mode switching while maintaining coding adaptability

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent dynamically selects the coefficient scan path (vertical or horizontal) based on the transform unit aspect ratio, but maintains a consistent coding group structure within each path. This dynamic selection reduces processing overhead by avoiding frequent mode switching while still adapting to different block shapes

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP3484149B1Method and apparatus for transform coefficient coding of non-square blocks
Publication Date: 2020.11.11 MEDIATEK SINGAPORE PTE LTD
  • EP3484149B1 patent drawingFigure 1
  • EP3484149B1 patent drawingFigure 2
  • EP3484149B1 patent drawingFigure 3~5

AI summary

A method and apparatus for transform coefficient coding of image and video data for a video encoder or decoder are disclosed. The method determines a CG (coding group) size adaptively based on the current TU. The current TU is divided into one or more current CGs (coding groups) according to the CG size. Bins associated with the coefficients of the current TU are then encoded or decoded according to a selected CG scan path through the current TU and one or more coefficient scan paths within each CG. If TU width is larger or smaller than TU height for the current TU, then CG width is selected to be larger or smaller than CG height accordingly. According to another method, all bypass-coded bins associated with coefficients in the current TU are grouped to improve the parsing throughput rate.