Video Block Merge Candidate Ranking for Faster Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional video coding methods waste time and resources by estimating the difference degree between block units and reference blocks for each merge candidate, leading to inefficient encoding and decoding processes.
Innovation Solution
A method and device that utilize template matching to arrange and select merge candidates, grouping them into subgroups with limited numbers, and ordering them based on cost values to efficiently predict block units.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If detailed estimation is performed for each merge candidate to determine the difference degree between the block unit and reference block, then the prediction accuracy is improved, but the encoding and decoding time increases significantly
Solution Approach 1:
The patent segments the merge candidate list into multiple subgroups, where each subgroup contains a limited number of merge candidates. This segmentation allows the encoder and decoder to process fewer candidates in detail while maintaining prediction accuracy by selecting from diverse subgroups.
Solution Approach 2:
The patent performs preliminary arrangement of merge candidates into subgroups based on template matching cost values before the actual prediction process. This preliminary organization enables faster selection during encoding and decoding by pre-sorting candidates according to their matching quality.
2Measurement precision
If the number of merge candidates is increased to improve prediction quality, then the prediction accuracy is improved, but the computational complexity increases
Solution Approach 1:
The patent divides the large set of merge candidates into multiple smaller subgroups, each containing a limited number of candidates. This segmentation reduces the computational burden by allowing parallel or selective processing of subgroups rather than evaluating all candidates individually.
Solution Approach 2:
The patent performs detailed estimation only for a limited number of top-ranked merge candidates from each subgroup rather than all candidates. This partial action approach maintains prediction quality by focusing computational resources on the most promising candidates while ignoring less relevant ones.
Data Source
AI summary
A method of encoding video data is provided. The method determines a block unit of an image frame of the video data. The method determines, for the block unit, a first merge candidate list including multiple merge candidates, and identifies multiple merge subgroups from the first merge candidate list. The method determines multiple first cost values each corresponding to one of the merge candidates, and determines an arrangement of the merge candidates in each of the merge subgroups based on the first cost values. The method determines a second merge candidate list by selecting, from each of the merge subgroups, a first K merge candidates of the merge candidates ordered based on the arrangements. Then, the method selects one of the merge candidates in the second merge candidate list to predict the block unit and encodes one or more merge indices into a bitstream based on the selected merge candidate.