Video Coding Quantization Prediction Across Block Boundaries
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional moving picture coding technologies face inefficiencies in coding quantization parameters, particularly when block sizes are small and neighboring blocks have different characteristics, leading to increased coding amounts due to large differences in quantization parameters and separate processing sequences.
Innovation Solution
A moving picture coding device that further divides blocks into smaller units, calculates quantization parameters, and uses neighboring blocks to derive predictive quantization parameters, generating differential quantization parameters to reduce coding amounts by optimizing the prediction of quantization parameters across block boundaries.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If differential coding of quantization parameter is performed using the quantization parameter of the previous block in the coding sequence, then the information amount of the quantization parameter is suppressed, but when the block size is small and neighboring blocks have different characteristics, the processing sequence between slices is separated and the coding amount cannot be reduced
Solution Approach 1:
The patent applies local quality by selecting different prediction sources based on the local characteristics of neighboring blocks. When blocks are divided into small sizes and neighboring blocks have different characteristics (e.g., different slice boundaries or different picture regions), the patent selectively uses either the left neighboring block or the upper neighboring block as the prediction source, rather than universally using the previous block in coding sequence. This localized adaptation ensures that the prediction is always based on the most similar local region, maintaining coding efficiency even when global coding sequence separation occurs.
2Loss of information
If the quantization parameter of the left block is used for prediction, then the coding amount is reduced in general cases, but when the characteristics of pictures inside the coding target block and the left block are different, the difference between quantization parameters becomes large and the coding amount increases
Solution Approach 1:
The patent implements dynamics by making the prediction source selection adaptive and changeable based on local characteristics. Instead of statically using the left block for all predictions, the system dynamically determines whether to use the left block or the upper block as the prediction source by comparing local characteristics (such as activity levels, gradient directions, or slice boundaries). This dynamic adaptation allows the prediction mechanism to respond to local variations in picture content, maintaining high prediction accuracy across diverse block characteristics.
3Manufacturing precision
If blocks are further divided into smaller units to improve picture quality control, then the quantization can be optimized for each small block, but the processing complexity and the difficulty of determining appropriate prediction sources increases
Solution Approach 1:
The patent applies segmentation by dividing the prediction source selection process into distinct cases based on block position and characteristics. Rather than implementing a complex continuous optimization algorithm, the patent segments the problem into discrete scenarios (e.g., blocks at slice boundaries, blocks in different picture regions, blocks with different activity levels) and applies simple decision rules to each segment. This segmented approach maintains low processing complexity while enabling precise quality control for each small block through appropriate local prediction.
Data Source
AI summary
In a moving picture coding device that further divides a first block acquired by dividing each picture of a moving picture into a predetermined size into one or a plurality of second blocks and codes the moving picture in units of blocks, a quantization parameter calculation unit calculates a quantization parameter of the second block. A predictive quantization parameter deriving unit derives a predictive quantization parameter of the second block by using the quantization parameter of one or a plurality of third blocks that are neighboring to the second block. The predictive quantization parameter deriving unit derives the predictive quantization parameter of the second block by using a quantization parameter of a fourth block coded before the second block in a case where the third block neighboring to the second block is located at a position beyond a boundary of the first block.