Moving Image Coding Using Parallelism Index for Block Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current block-matching techniques for moving image compression struggle to accurately identify conditions of effective resemblance and total extraneity between blocks, leading to inefficient coding and high output bits due to the asymmetrical confidence intervals of measurement indices like SAD and SATD, which do not account for error distribution at a spatial level.
Innovation Solution
A method that calculates an index of parallelism between surfaces by analyzing the distribution of differences pixel by pixel, using a 3D space representation to determine the degree of parallelism and choose the best predictor block, even for blocks with different average luminance, thereby optimizing coding by considering the distribution of differences.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If classic block-matching techniques using indices like SAD or SATD are used to select predictor blocks, then the coding process is simple and fast, but the reliability of block matching is insufficient due to asymmetrical confidence intervals and inability to account for error distribution at spatial level
Solution Approach 1:
The patent transitions from analyzing scalar difference values to analyzing the spatial distribution of errors across the block. By examining error distribution patterns in the spatial domain, the method adds a dimensional aspect to the evaluation, enabling more reliable block matching while maintaining computational feasibility through structured analysis approaches.
2Measurement precision
If exhaustive search or recursive search methods are used to find the best predictor block, then the accuracy of motion estimation is improved, but the computational burden and processing time increase significantly
Solution Approach 1:
The patent performs preliminary analysis of error distribution characteristics before final predictor block selection. By evaluating spatial error patterns in advance and identifying promising candidate blocks based on these patterns, the method narrows down the search space, allowing accurate motion estimation without requiring exhaustive search of all possible blocks.
Solution Approach 2:
The patent applies different evaluation criteria to different regions within the block based on local error distribution characteristics. By analyzing spatial variations in error patterns and adapting the selection process to local conditions, the method achieves high accuracy while reducing the overall computational burden compared to uniform exhaustive search.
3Loss of substance
If the number of bits for quantization is limited to reduce data量, then the compression ratio is improved, but the loss of information increases due to insufficient bits to represent the difference signal accurately
Solution Approach 1:
The patent extracts and utilizes the spatial distribution pattern of errors as additional information. By analyzing how errors are distributed across the block spatially, the method extracts meaningful structural information that can be leveraged for more efficient coding, allowing better reconstruction with fewer bits by capturing the essential spatial characteristics rather than all raw difference values.
Data Source
AI summary
A moving-image signal, such as typically a luminance signal organized in blocks of pixels is coded via a technique that envisages the steps of: comparing a block to be coded with a plurality of candidate prediction blocks; determining, for each candidate prediction block, a respective value of an index representing the difference between the block to be coded and each candidate prediction block; and choosing between the candidate prediction blocks, as a function of the respective value of the aforesaid index, a prediction block to be used for coding of the block to be coded. The signal is sampled pixel by pixel on the block to be coded and on the plurality of candidate prediction blocks, thus generating respective surfaces representing the pattern of the signal in the block to be coded and in the candidate prediction blocks. Chosen as an index is an index of the parallelism between the aforesaid respective surfaces.