Intra-prediction Type Classification for Video Processing Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing intra-prediction methods in video data processing, such as in the H.264 standard, require empirical testing of all possible predictive modes for each macro-block, leading to delays and inefficiencies due to the need to determine the optimal predictive mode among numerous options.
Innovation Solution
A method and system that classify pixel blocks into edge classes using an edge classifier, allowing for the identification of homogeneous and heterogeneous blocks, which enables the determination of optimal intra-prediction types without empirical testing of all possible modes, by assigning directions and selecting macro-block types based on region characteristics like flat, high detail, or moderate detail regions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all possible predictive modes are empirically tested to determine the optimal intra-prediction type, then prediction accuracy is improved, but processing time and computational complexity increase
Solution Approach 1:
The patent segments the image into different regions (flat regions, edge regions, texture regions) and applies different prediction strategies to each region. For flat regions, a simplified prediction method is used that requires minimal testing, while for edge and texture regions, more comprehensive testing is performed. This segmentation allows the system to achieve good overall prediction accuracy without having to perform exhaustive testing on every block, thus reducing processing time while maintaining prediction quality.
Solution Approach 2:
The patent applies different levels of prediction mode testing to different local regions based on their characteristics. In flat regions where pixel values are similar, only a few prediction modes are tested. In edge and texture regions where more variation exists, more prediction modes are tested. This local quality approach ensures that computational resources are focused where they are most needed, improving accuracy in complex regions while reducing unnecessary processing in simple regions.
2Reliability
If all possible predictive modes are empirically tested for each macro-block, then optimal prediction mode selection is improved, but device complexity and processing overhead increase
Solution Approach 1:
The patent performs preliminary classification of image regions before conducting prediction mode testing. By first identifying whether a region is flat, edge, or texture, the system can pre-determine which prediction modes are most likely to be optimal for that region. This preliminary action eliminates the need to test all possible prediction modes uniformly across the entire image, significantly reducing processing overhead while maintaining reliable mode selection through targeted testing in each region type.
Solution Approach 2:
The patent applies partial testing of prediction modes rather than exhaustive testing. For flat regions, only a subset of prediction modes is tested (partial action), while for edge and texture regions, a more comprehensive but still selective testing approach is used. This avoids the excessive action of testing all 184 possible modes in every region, reducing device complexity and processing overhead while maintaining sufficient reliability for optimal mode selection.
3Measurement precision
If exhaustive empirical testing is performed to determine intra-prediction type, then prediction precision is improved, but productivity decreases
Solution Approach 1:
The patent divides the processing task into segments based on region characteristics. Flat regions are processed with a simplified prediction approach that achieves sufficient precision with minimal testing, while edge and texture regions receive more intensive processing. This segmentation allows the system to maintain good prediction precision across the entire image while significantly improving overall processing efficiency by avoiding exhaustive testing in regions where it is unnecessary.
Solution Approach 2:
The patent changes the testing parameters (number of modes tested, testing depth) based on region characteristics. In flat regions, fewer modes are tested and simpler metrics are used, while in edge and texture regions, more modes are tested with more comprehensive evaluation. This dynamic parameter adjustment allows the system to maintain prediction precision where needed while improving productivity by reducing processing in regions where high precision is less critical.
Data Source
AI summary
The method, system, and apparatus of source statistics based intra prediction type is disclosed. In one embodiment, a method includes classifying a four-pixel square block in an edge class (e.g., may include a DC edge class, a vertical edge class, a horizontal edge class, a diagonal edge class, and/or a planar edge class) based on an edge classifier, classifying an eight-pixel square block having the four-pixel square block and other four-pixel square blocks as a homogenous class if the four-pixel square block and the other four-pixel square blocks of the eight-pixel square block belong to the edge class, assigning a direction to the edge class of the eight-pixel square block, and determining an optimal intra-prediction type through the classification such that empirical testing of all possible ones of the edge class and the direction is avoided when the homogenous class is identified.