Color Index Prediction Using Pattern Relation Table

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current color index coding techniques in High Efficiency Video Coding (HEVC) extensions, such as range extensions (RExt), face limitations in coding efficiency due to the complexity of predicting and transmitting color palettes across blocks, especially in screen content materials with varying color formats and bit-depths.

Innovation Solution

The introduction of a 'Copy by Pattern Search' mode in color index coding, where predictors are derived based on previous coded pixels using a pattern relation table (PRT) to efficiently encode or decode pixel values, allowing for adaptive prediction and reduced redundancy in encoding processes.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If color index coding is used to represent blocks of pixels using indices to the palette, then coding efficiency is improved for screen content materials, but the complexity of predicting and transmitting color palettes across blocks increases

Engineering Contradiction:
Improvecoding efficiencyVSAvoidpalette prediction complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent applies copying by deriving predictors from previously coded pixels within the current block using a pattern relation table (PRT). Instead of transmitting or predicting palettes from external sources, the method copies color information from already-coded pixels in the same block, exploiting spatial redundancy to reduce bitrate while maintaining coding efficiency.

Inventive Principle:
Principle #26Copying

2Quantity of substance

If palette prediction is performed from neighboring blocks, then bitrate is reduced, but the adaptability to varying color formats and bit-depths decreases

Engineering Contradiction:
ImprovebitrateVSAvoidcolor format adaptability
Core Design Contradiction:
Quantity of substanceVSAdaptability or versatility

Solution Approach 1:

The patent implements self-service by having the current block predict its own color indices from previously coded pixels within itself using the PRT. This self-contained prediction approach eliminates dependency on external palette transmission, automatically adapting to the specific color format and bit-depth of the current block while reducing bitrate.

Inventive Principle:
Principle #25Self-service

3Ease of manufacture

If conventional color index coding is used, then implementation is simpler, but parsing throughput and coding performance are lower

Engineering Contradiction:
Improveimplementation simplicityVSAvoidparsing throughput
Core Design Contradiction:
Ease of manufactureVSProductivity

Solution Approach 1:

The patent applies preliminary action by pre-computing and storing the pattern relation table (PRT) during the encoding process. The PRT captures the relationships between previously coded pixels and current pixels, allowing for faster decoding and parsing throughput without significantly increasing implementation complexity, as the table is built incrementally during normal encoding.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentUS10484696B2Method and apparatus for color index prediction
Publication Date: 2019.11.19 MEDIATEK INC
  • US10484696B2 patent drawing
  • US10484696B2 patent drawing
  • US10484696B2 patent drawing

AI summary

A method and apparatus for coding a block of video data using index or pixel value prediction including a copy-by-pattern-search mode are disclosed. According to the present invention, a current search pattern is determined based on one or more previous coded pixels. One or more predictors are derived according to the current search pattern for a current index or pixel value of a current pixel in the current block. Encoding or decoding is then applied to one or more following indices or pixel values of one or more following pixels including the current pixel using the predictors. In a simplest case, the one or more previous coded pixels correspond to a single previous coded pixel at the left side of the current pixel and the one or more following pixels contain only the current pixel.