Quadratic Image Transform Coding for Intra-Predicted Subblocks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increasing demand for high-resolution and high-quality images/videos, particularly in immersive media formats, leads to higher transmission and storage costs due to increased bit amounts, necessitating a more efficient image/video compression technique.
Innovation Solution
An image coding method and apparatus that enhances coding efficiency by applying a non-separable secondary transform with a transform kernel matrix optimized for intra prediction modes, partitioning blocks into subblocks, and using a unified transform kernel matrix across subblocks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional transform methods are used for high-resolution images, then implementation is simple, but compression efficiency is insufficient
Solution Approach 1:
The transform processing is segmented into a primary transform and a secondary transform. The primary transform processes the entire block, while the secondary transform is selectively applied to specific regions (e.g., regions with non-zero coefficients) based on coefficient distribution patterns. This segmentation allows the system to achieve higher compression efficiency through adaptive processing while maintaining manageable complexity by applying complex operations only where necessary.
Solution Approach 2:
The transform methodology dynamically adapts based on the characteristics of the input data. The system determines whether to apply a secondary transform by analyzing coefficient distribution patterns, and selects different transform types (e.g., DST-II, DCT-II) based on prediction modes and coefficient locations. This dynamic adaptation enables the system to optimize compression efficiency for different image regions and content types without requiring complex processing for all cases.
2Loss of energy
If high-resolution images are transmitted without advanced compression, then image quality is maintained, but transmission cost increases
Solution Approach 1:
The system changes transform parameters adaptively based on image characteristics. Different transform types are selected based on prediction modes (e.g., directional prediction modes trigger specific secondary transforms), and transform coefficients are processed differently based on their distribution patterns. This parameter adaptation enables more efficient energy compaction, reducing the number of bits needed to represent the same image quality, thereby lowering transmission costs while maintaining image fidelity.
3Productivity
If a single transform method is used for all blocks, then processing is uniform, but coding efficiency for specific regions is insufficient
Solution Approach 1:
The system applies different transform methods to different regions of the image based on local characteristics. Secondary transforms are applied selectively to regions containing non-zero coefficients or regions with specific prediction modes, while other regions use only the primary transform. This local adaptation improves coding efficiency by matching the transform method to the local image content, while the region-based approach keeps the complexity manageable by avoiding universal complex processing.
Data Source
AI summary
An image decoding method according to the present document comprises the steps of: receiving a quantized transform coefficient for a target block and a transform index for non-separable quadratic transform; deriving transform coefficients by inversely quantizing the quantized transform coefficient; deriving corrected transform coefficients on the basis of a transform kernel matrix in a predetermined transform set indicated by the transform index; and deriving residual samples for the target block on the basis of inverse linear transform for the corrected transform coefficients, wherein when the target block is divided into a predetermined number of sub-blocks and is coded by intra prediction, the corrected transform coefficients are derived in units of the sub-blocks, and the transform index is received for the target block.