Video Encoding Mixed Prediction Mode Selection

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current video encoding techniques face high computational complexity due to the large number of available prediction modes, which increases computational overhead, and existing methods to improve rate distortion performance by adding more modes exacerbate this issue.

Innovation Solution

The method involves determining a primary prediction mode and a companion prediction mode, generating a mixed prediction by combining them, and using accuracy maps to identify effective prediction modes, with optional adaptive updating of lookup tables based on decoding results.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If more intra-prediction modes are added to improve rate distortion performance, then the coding efficiency is improved, but the computational complexity increases excessively

Engineering Contradiction:
Improverate distortion performanceVSAvoidcomputational complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent segments the prediction mode selection process into two distinct stages: (1) selecting a primary prediction mode from a reduced set of candidate modes, and (2) determining a companion prediction mode based on the primary mode selection. This segmentation reduces the computational burden by avoiding exhaustive evaluation of all possible mode combinations while still achieving improved rate distortion performance through the companion mode mechanism.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary action by pre-defining a limited set of candidate primary prediction modes and establishing predetermined relationships between primary modes and their corresponding companion modes. This preliminary structuring allows the encoder/decoder to quickly identify mode pairs without performing complex real-time calculations, thereby reducing computational complexity while maintaining coding efficiency.

Inventive Principle:
Principle #10Preliminary action

2Measurement precision

If a brute-force implementation of MDL minimization is used to select prediction modes, then the optimal mode selection is achieved, but the computational overhead becomes excessive

Engineering Contradiction:
Improvemode selection accuracyVSAvoidcomputational overhead
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent extracts the essential function of MDL minimization by focusing only on selecting a primary prediction mode from a reduced candidate set, rather than performing exhaustive MDL calculations for all possible modes. The companion prediction mode is then determined based on this primary selection, effectively extracting the core benefit of MDL minimization while avoiding its computational burden.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent applies partial action by implementing a simplified two-stage mode selection process that performs MDL minimization only for the primary prediction mode selection, rather than for all possible mode combinations. This partial application of the minimization principle achieves sufficient mode selection accuracy while dramatically reducing computational overhead compared to a complete brute-force approach.

Inventive Principle:
Principle #16Partial or excessive action

3Productivity

If the number of available prediction modes is increased from 35 to further improve performance, then the coding efficiency improves, but the computational complexity increases even further

Engineering Contradiction:
Improvecoding efficiencyVSAvoidcomputational overhead
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent introduces dynamics by making the companion prediction mode selection adaptive based on the primary prediction mode chosen. Rather than evaluating all modes statically, the system dynamically determines which companion mode to use based on the primary mode selection and predefined relationships, allowing the effective number of evaluated modes to adapt to the specific coding situation and reducing overall computational complexity.

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP3152901B1Apparatus and method to support encoding and decoding video data
Publication Date: 2020.10.28 BLACKBERRY LTD
  • EP3152901B1 patent drawingFigure 1
  • EP3152901B1 patent drawingFigure 2~4
  • EP3152901B1 patent drawing

AI summary

To support encoding a control circuit determines a primary prediction mode to employ when encoding a block of video data and then determines a corresponding companion prediction mode to correspond to the primary prediction mode. The control circuit can then generate a mixed prediction for the block of video data by mixing a prediction formed using the primary prediction mode with a prediction formed using the companion prediction mode. By one approach the control circuit determines the corresponding companion prediction mode by accessing a companion prediction mode lookup table. In such a case, if desired, upon generating the mixed prediction for the block of video data the control circuit can also automatically update the companion prediction mode lookup table. By one approach a corresponding decoder infers the employed companion prediction mode (or modes).