Mode-Selected Secondary Transforms for Video Signal Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video signal processing methods lack efficiency in coding and signaling of motion information sets, requiring improved techniques for enhanced compression and decoding processes.
Innovation Solution
A video signal processing method that applies a secondary inverse transform to current blocks based on intra prediction modes, using predefined secondary transform kernel sets and a fixed scan order, with optional allocation of inverse quantized transform coefficients in an up-right diagonal scan order, and splitting blocks when necessary.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a secondary inverse transform is applied to improve coding efficiency, then compression performance is improved, but processing complexity increases
Solution Approach 1:
The transform processing is segmented into a primary inverse transform and an optional secondary inverse transform. The secondary transform is applied selectively to specific blocks based on prediction mode and block size conditions, rather than uniformly to all blocks. This segmentation allows the system to achieve improved coding efficiency for suitable blocks while avoiding unnecessary complexity for blocks where the secondary transform would not be beneficial.
Solution Approach 2:
The application of the secondary inverse transform is made dynamic through conditional logic that evaluates block characteristics (prediction mode, block size) before applying the transform. The system adapts its processing complexity based on the specific block being decoded, applying the secondary transform only when conditions are met, thereby optimizing the trade-off between coding efficiency and processing complexity on a per-block basis.
2Measurement precision
If multiple transform kernel sets are used to improve transform accuracy, then processing precision is improved, but device complexity increases
Solution Approach 1:
Different transform kernel sets are assigned to different prediction modes (e.g., planar, DC, angular modes). Each prediction mode has its own optimized kernel set that is locally adapted to the characteristics of that mode. This local quality approach ensures that the most appropriate kernels are used for each specific prediction type, improving transform accuracy without requiring the system to evaluate all possible kernels for every block.
Solution Approach 2:
The system changes the transform kernel parameters based on the prediction mode of the block. By linking kernel selection to prediction mode parameters, the system automatically selects the appropriate kernel set without requiring complex real-time analysis. This parameter-based selection simplifies the decision process while maintaining high transform accuracy through mode-specific optimization.
3Measurement precision
If transform coefficients are allocated using variable scan orders to improve reconstruction accuracy, then processing precision is improved, but processing time increases
Solution Approach 1:
The scan orders for allocating transform coefficients are predetermined and stored as lookup tables for different block sizes and transform types. Rather than calculating scan orders dynamically during decoding, the system uses pre-computed scan patterns that are selected based on block characteristics. This preliminary action significantly reduces processing time while maintaining the accuracy benefits of optimized scan orders.
Solution Approach 2:
The scan order parameters are changed based on block size and transform type, with different predefined scan patterns used for different scenarios. This parameter-based selection allows the system to quickly adapt to different block configurations without complex real-time optimization, balancing reconstruction accuracy with processing efficiency through pre-established parameter mappings.
Data Source
AI summary
A video signal processor is configured to obtain a secondary transform kernel for a current block based on an intra prediction mode of the current block to which a secondary transform is applied, to obtain a secondary inverse transformed block by performing a secondary inverse transform on a top-left specific region of the current block using the secondary transform kernel, wherein the secondary inverse transform is an inverse transform of the secondary transform, and the secondary transform is a low frequency non-separable transform, to obtain a residual block of the current block by performing a primary inverse transform on the secondary inverse transformed block, wherein one or more coefficients of the top-left specific region of the current block are derived in a preset scan order and the preset scan order is a 4×4 up-right di agonal scan order regardless of a size of the current block.