Triangular Merge Mode Harmonization With Weighted Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face challenges in efficiently compressing video data while maintaining high picture quality, particularly in scenarios with limited bandwidth and memory resources.
Innovation Solution
The proposed method involves encoding and decoding video sequences using weighted prediction parameters that combine fade weighting parameters and blending weighting parameters, with the values of fade weighting parameters determined by a reference index value and a reference picture list, and blending weighting parameters determined by the position of a predicted sample in a predicted block.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If non-rectangular prediction units are used with slice-level weighted prediction, then prediction accuracy for fades is improved, but signaling overhead increases
Solution Approach 1:
The patent applies local quality by enabling non-rectangular prediction units and slice-level weighted prediction only in specific regions where fades are detected, rather than applying them globally. The encoder determines fade regions and restricts the use of advanced prediction modes to those areas, thereby improving prediction accuracy where needed while minimizing signaling overhead in other regions.
Solution Approach 2:
The patent segments the picture into slices and further segments fade regions within slices. By processing different slices independently and applying non-rectangular prediction units only to identified fade regions, the patent reduces the overall signaling overhead while maintaining prediction accuracy in critical areas.
2Productivity
If non-rectangular prediction units are enabled for slice-level weighted prediction, then compression efficiency is improved, but device complexity increases
Solution Approach 1:
The patent performs preliminary detection of fade regions before applying non-rectangular prediction units. By identifying fade regions in advance and preparing the necessary prediction data structures beforehand, the patent simplifies the overall processing complexity while maintaining high compression efficiency through targeted application of advanced prediction modes.
Solution Approach 2:
The patent applies non-rectangular prediction units partially, only to regions identified as containing fades, rather than applying them to the entire picture. This selective application improves compression efficiency for fade regions while avoiding the excessive processing complexity that would result from global application.
3Reliability
If reference pictures are repeatedly transmitted for weighted prediction, then prediction reliability is improved, but bandwidth consumption increases
Solution Approach 1:
The patent makes reference pictures universal by storing them in a reference picture list that can be reused across multiple prediction units and slices. Instead of repeatedly transmitting the same reference picture data, the encoder transmits reference pictures once and reuses them through indexing, thereby maintaining prediction reliability while reducing bandwidth consumption.
Solution Approach 2:
The patent uses copying by creating reference picture lists that contain indices to previously transmitted reference pictures. Rather than transmitting duplicate reference picture data multiple times, the encoder transmits copies (indices) that reference the original data, reducing bandwidth consumption while maintaining prediction reliability through accurate reference access.
Data Source
AI summary
The present disclosure relates to moving picture processing, and in particular, to cases where non-rectangular partitioning modes are used for inter-prediction in a combination with weighted prediction (WP) for coding fades. Non-rectangular modes refer to those inter-prediction modes, in which case a current block is partitioned/sliced in a non-rectangular way (slices). Such modes may be triangular (TMP) or geometric (GEO) prediction. The aim of the disclosure is to harmonize non-rectangular partitioning modes with WP by way of disabling TMP/GEO when WP is applied. This may be accomplished by checking whether all reference pictures for a current slice are disabled, and if this is the case to disable the slice-level WP for non-rectangular prediction units. Whether the reference pictures are disabled may be determined based on conditions applied on set values, for example, of luma/chroma flags, which refer to the reference pictures.