Video Decoding Merge Candidate List and Quantization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The H.264 video coding standard faces inefficiencies in motion vector coding and quantization parameter adjustment, particularly when motion differs from neighboring blocks, leading to decreased coding efficiency and increased bit requirements.
Innovation Solution
A method for decoding video data that constructs a merge candidate list using spatial and temporal candidates, derives motion information with a merge index, generates prediction and residual blocks through inverse quantization and transformation, and adjusts quantization parameters per unit to improve coding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If motion vector predictor is generated using neighboring blocks in H.264, then coding efficiency is improved when motion is similar, but coding efficiency degrades when motion differs from neighboring blocks
Solution Approach 1:
The patent applies dynamics by making the motion prediction system adaptive rather than static. It dynamically selects between spatial prediction (using neighboring blocks) and temporal prediction (using co-located blocks from reference pictures) based on the actual motion characteristics. This dynamic adaptation allows the system to maintain high coding efficiency whether motion is similar to neighbors or different, resolving the contradiction between efficiency and adaptability.
Solution Approach 2:
The patent changes the parameter selection strategy by introducing multiple prediction modes (spatial vs. temporal) and allowing the system to switch between them. Instead of fixed parameter usage, the system evaluates motion characteristics and selects the appropriate prediction mode, thereby maintaining coding efficiency across varying motion patterns and resolving the adaptability issue.
2Manufacturing precision
If quantization parameter is adjusted per macroblock in H.264, then coding precision is improved, but the amount of coding bits increases
Solution Approach 1:
The patent applies segmentation by dividing the picture into different hierarchical levels (picture level, slice level, and macroblock level) for quantization parameter control. Instead of uniformly adjusting QP per macroblock, the system segments the control granularity, allowing coarse adjustment at higher levels and fine adjustment only where necessary, thereby maintaining precision while reducing overall bit consumption.
Solution Approach 2:
The patent implements local quality by allowing different quantization precision levels in different regions. Important regions (such as regions with high visual importance or motion activity) receive finer quantization control, while less important regions use coarser control. This selective approach maintains necessary precision while minimizing the total number of bits required for QP signaling.
3Adaptability or versatility
If various sizes of coding units are used for inter prediction, then adaptability is improved, but correlation between quantization parameter and previous quantization parameter becomes weaker
Solution Approach 1:
The patent resolves this contradiction by introducing a new dimension of control through separate QP adjustment mechanisms for different coding unit sizes. Instead of relying solely on temporal correlation which weakens with size variation, the system establishes size-specific QP relationships, allowing each coding unit size to have its own correlation characteristics. This dimensional approach maintains reliability while preserving adaptability to various block sizes.
Data Source
AI summary
A method of decoding video data in merge mode can include constructing a merge candidate list using available spatial and temporal merge candidates, deriving motion information using a merge index and the merge candidate list, generating a prediction block using the motion information, generating a residual block by inverse-quantizing a quantized block using a quantization parameter and a quantization matrix and by inverse-transforming the inverse quantized block, and generating a reconstructed block using the residual block and the prediction block. The quantization parameter is generated per quantization unit and a minimum size of the quantization unit is adjusted per picture, and the quantization parameter is generated using a quantization parameter predictor and a differential quantization parameter. Also, the differential quantization parameter is generated by restoring a bin string indicating an absolute value of the differential quantization parameter and a bin indicating a sign of the differential quantization parameter.