Reduced Secondary Transform for High-Quality Video Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increasing demand for high-resolution and high-quality images/videos, including immersive media, necessitates a highly efficient image/video compression technique to reduce transmission and storage costs.
Innovation Solution
An image coding method and apparatus utilizing a reduced secondary transform (RST) with optimized transform kernel matrices and transform sets based on intra prediction modes to enhance coding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high-resolution and high-quality images/videos are transmitted or stored using conventional methods, then image quality is maintained, but transmission cost and storage cost increase
Solution Approach 1:
The patent applies transform parameters (DCT, DST types) and quantization parameters dynamically based on prediction modes and block characteristics. This optimization of transformation parameters improves compression efficiency by adapting the transform type to the specific characteristics of each block, thereby reducing the bitrate required to maintain high image quality without sacrificing fidelity
Solution Approach 2:
The patent introduces dynamic selection of transform types (DCT-2, DST-7, DCT-8) and secondary transforms based on intra prediction modes, block sizes, and gradient characteristics. This dynamic adaptation allows the system to optimize compression for each specific block rather than using a fixed transform, achieving better compression ratios while maintaining high quality
2Ease of manufacture
If conventional transform methods are used for image coding, then implementation is simple, but transform efficiency and coding efficiency are insufficient
Solution Approach 1:
The patent divides the transform process into multiple stages: primary transform (DCT/DST), secondary transform (NST variants), and quantization. Each stage can be independently optimized and selected based on block characteristics. This segmentation allows the system to apply complex multi-stage transforms when needed while maintaining the option to use simpler single-stage transforms for other blocks
Solution Approach 2:
The patent creates a universal transform framework that can handle multiple transform types (DCT-2, DST-7, DCT-8) and secondary transforms (NST, RST) through a unified interface. The same basic architecture supports different transform kernels and configurations, allowing flexible adaptation to various block types without requiring separate implementation paths
3Productivity
If a reduced secondary transform (RST) is applied to transform coefficients, then coding efficiency increases, but computational complexity increases
Solution Approach 1:
The patent applies reduced secondary transform only to specific frequency bands or coefficient regions where it provides the most benefit, rather than to all coefficients. The transform is selectively applied based on coefficient magnitude thresholds and frequency characteristics, achieving improved coding efficiency for the most significant coefficients while limiting the computational overhead to only where necessary
Solution Approach 2:
The patent applies different transform strategies to different regions of the transform coefficient matrix. The reduced secondary transform is applied selectively to low-frequency or high-magnitude coefficients where it provides maximum benefit, while high-frequency or low-magnitude coefficients may use simpler processing. This localized application optimizes the balance between coding efficiency and computational complexity
Data Source
AI summary
A video decoding method according to the present document comprises the steps of: deriving transform coefficients by performing de-quantization on the basis of quantized transform coefficients for a target block; deriving modified transform coefficients on the basis of an inverse reduced secondary transform (RST) of the transform coefficients; and generating a reconstructed picture on the basis of residual samples for the target block on the basis of an inverse primary transform of the modified transform coefficients, wherein an inverse RST using a transform kernel matrix is performed on transform coefficients of an upper-left 4×4 region of an 8×8 region of the target block, and modified transform coefficients of the upper-left 4×4 region, an upper-right 4×4 region, and a lower-left 4×4 region of the 8×8 region are derived through the inverse RST.