Multi-hypothesis Video Coding Block Partitioning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding technologies face inefficiencies in resource usage and computational complexity, particularly in HEVC's Multi-hypothesis (MH) mode, which can be improved by simplifying the prediction process and reducing system resource requirements.
Innovation Solution
A method is introduced to partition a current block into two partitions, with candidates derived from a regular Merge mode candidate list, allowing for blended predictions that reduce memory access bandwidth and computational complexity by applying a Multi-hypothesis prediction mode that combines existing and additional Inter-indexed predictions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If Multi-hypothesis mode is implemented to improve prediction accuracy, then coding efficiency is improved, but computational complexity and system resource requirements increase
Solution Approach 1:
The current block is divided into two partitions, with each partition using different candidate derivation methods (one using regular Merge mode candidate list, another using additional Inter-indexed predictions). This segmentation allows the system to achieve multi-hypothesis prediction benefits while controlling complexity by processing different regions with different methods.
Solution Approach 2:
The patent makes the existing Merge mode candidate list serve multiple functions: it is used for both regular Merge mode coding and as the basis for Multi-hypothesis mode candidate derivation. This multi-functionality eliminates the need for separate candidate lists, reducing memory bandwidth requirements while maintaining prediction accuracy.
2Measurement precision
If Multi-hypothesis mode uses separate candidate lists for different partitions, then prediction accuracy is improved, but memory access bandwidth increases
Solution Approach 1:
The regular Merge mode candidate list is designed to serve dual purposes: it provides candidates for standard Merge mode operation and simultaneously serves as the foundation for Multi-hypothesis mode candidate derivation. This eliminates the need to maintain separate candidate lists in memory, significantly reducing memory access bandwidth requirements while preserving the ability to generate accurate predictions through multiple hypotheses.
3Measurement precision
If traditional Multi-hypothesis mode is used with additional candidate lists, then prediction accuracy is improved, but device complexity increases
Solution Approach 1:
The block is segmented into two partitions that can be processed differently, allowing the system to apply complex multi-hypothesis processing only where necessary while using simpler methods elsewhere. This selective application reduces overall system resource requirements while maintaining prediction accuracy in critical regions.
Solution Approach 2:
The existing Merge mode infrastructure is made multi-functional to support both traditional and multi-hypothesis operations. By reusing the established candidate list structure and derivation processes for multiple purposes, the patent avoids duplicating system resources while enabling enhanced prediction capabilities.
Data Source
AI summary
A method and apparatus of prediction for video coding using MH (Multiple Hypothesis) mode are disclosed. According to this method, a block is partitioned into a first partition and a second partition. A first candidate and a second candidate are derived for the first and second partitions respectively. At least one of the first candidate and the second candidate is derived using a candidate list derived for a regular Merge mode (or also called as normal Merge mode). An MH prediction generated by blending a first prediction corresponding to a first candidate and a second prediction corresponding to a second candidate, and the MH prediction is applied to a part of the current block.