Intra Prediction Direction Inheritance in Scalable Video Coding

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for intra-picture coding in spatial scalable video coding do not efficiently exploit the correlation between base and enhancement layers, leading to suboptimal coding efficiency and increased bitrate due to the need to encode intra-prediction mode as side information.

Innovation Solution

The introduction of the 'INTRA_DIRECT' method, which inherits intra prediction directions from the base layer and up-scales them for the enhancement layer, eliminating the need to encode intra-prediction mode and reducing side information, thereby improving coding efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If intra prediction mode is encoded as side information for each macroblock in the enhancement layer, then prediction accuracy is improved, but bitrate increases

Engineering Contradiction:
Improveprediction accuracyVSAvoidbitrate
Core Design Contradiction:
Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent copies the intra prediction mode from the base layer macroblock to the enhancement layer macroblock. Instead of encoding separate prediction modes for each enhancement layer macroblock, the system reuses the prediction mode information from the corresponding base layer, significantly reducing the side information bitrate while maintaining prediction accuracy through the copied mode data.

Inventive Principle:
Principle #26Copying

2Adaptability or versatility

If separate intra prediction is performed for base layer and enhancement layer, then layer independence is maintained, but coding efficiency decreases

Engineering Contradiction:
Improvelayer independenceVSAvoidcoding efficiency
Core Design Contradiction:
Adaptability or versatilityVSProductivity

Solution Approach 1:

The patent merges the intra prediction process between base layer and enhancement layer by performing prediction operations that utilize data from both layers. The enhancement layer macroblocks are predicted using a combination of base layer macroblock data and enhancement layer residual data, creating a unified prediction approach that improves coding efficiency while preserving layer structure.

Inventive Principle:
Principle #5Merging (Combining)

3Measurement precision

If base layer picture is upsampled using 6-tap filter for prediction, then prediction quality is improved, but computational complexity increases

Engineering Contradiction:
Improveprediction qualityVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent applies the computationally intensive 6-tap filter only in specific situations where high prediction quality is needed, rather than uniformly across all macroblocks. The system selectively applies upsampling based on local picture characteristics, maintaining high prediction quality where necessary while reducing overall computational complexity by avoiding unnecessary filtering operations.

Inventive Principle:
Principle #3Local quality

Data Source

PatentUS7847861B2Method and apparatus for encoding video pictures, and method and apparatus for decoding video pictures
Publication Date: 2010.12.07 INTERDIGITAL MADISON PATENT HLDG
  • US7847861B2 patent drawing
  • US7847861B2 patent drawing
  • US7847861B2 patent drawing

AI summary

The H.264 standard has introduced a new coding method for intra-coded pictures, which is named “intra prediction”. The basic idea of intra prediction is to predict a block using reconstructed neighbour pixels. Thus, only the residual information needs to be encoded. Scalable video coding adds some scalable extension to H.264, in terms of spatial, temporal and SNR scalability. An improved encoding method for scalable video comprises generating from a video picture a first and a second picture with higher resolution than the first picture, wherein each macroblock in the first picture has a plurality of corresponding macroblocks in the second picture, intra-coding the first picture on macroblock level, wherein for each macroblock a prediction direction is determined, generating a first residual picture, intra-coding the second picture on macroblock level, wherein for each macroblock the prediction direction of the corresponding macroblock of the first picture is reused, and generating a second residual picture.