Image Decoding with Secondary Transform and Adaptive Entropy Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image encoding and decoding technologies face challenges in efficiently compressing high-resolution or high-quality image content, particularly in terms of improving compression rates and optimizing prediction, motion compensation, and entropy coding processes.
Innovation Solution
The proposed method involves performing secondary inverse-transformation on sub-group units within transformation blocks, using adaptive probability models for entropy decoding, and employing pixel unit motion compensation to enhance the efficiency of image encoding and decoding processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional transformation and quantization methods are used, then the encoding process is simple, but the image compression rate is insufficient for high-resolution content
Solution Approach 1:
The transformation block is divided into multiple sub-groups, and secondary transformation is selectively applied to specific sub-groups based on their characteristics. This segmentation allows the patent to apply complex processing only where needed, improving compression rate while limiting overall complexity increase.
Solution Approach 2:
Different transformation strategies are applied to different regions of the image data. The patent performs secondary inverse-transformation on low-frequency components while skipping high-frequency components, applying local quality enhancement where it matters most for perceptual compression.
2Measurement precision
If a single probability model is used for entropy decoding, then the decoding process is fast, but the accuracy of probability estimation deteriorates
Solution Approach 1:
The patent dynamically switches between multiple probability models (first and second probability models) based on the characteristics of the data being decoded. This dynamic adaptation allows accurate probability estimation for different symbol types while maintaining reasonable decoding speed through conditional selection.
Solution Approach 2:
Multiple probability models are maintained to handle different types of syntax elements and data patterns. The first probability model handles certain cases while the second probability model handles others, creating a universal decoding framework that adapts to various input characteristics.
3Manufacturing precision
If motion compensation is performed at block level only, then the processing complexity is low, but the prediction accuracy deteriorates for regions with varying motion
Solution Approach 1:
The motion compensation process is segmented into two levels: block-level motion compensation for overall motion and pixel-level motion compensation for fine adjustments. This segmentation allows the patent to achieve high prediction accuracy in regions with varying motion while keeping the base complexity manageable through the block-level foundation.
Solution Approach 2:
Pixel-level motion compensation is applied selectively to enhance prediction accuracy in specific regions where block-level compensation is insufficient. This local enhancement approach improves overall prediction accuracy without applying complex processing uniformly across the entire image.
4Manufacturing precision
If secondary inverse-transformation is performed on all frequency components, then the reconstruction accuracy is improved, but the computational load increases significantly
Solution Approach 1:
Secondary inverse-transformation is applied selectively only to low-frequency components rather than all frequency components. This local quality approach improves reconstruction accuracy for the most perceptually important regions while significantly reducing computational load by skipping high-frequency components.
Solution Approach 2:
The patent performs secondary inverse-transformation on only a subset of components (low-frequency) rather than all components. This partial action provides sufficient reconstruction accuracy for perceptual purposes while avoiding the excessive computational cost of processing all frequency components.
Data Source
AI summary
An image decoding method and apparatus according to an embodiment may extract, from a bitstream, a quantization coefficient generated through core transformation, secondary transformation, and quantization; generate an inverse-quantization coefficient by performing inverse quantization on the quantization coefficient; generate a secondary inverse-transformation coefficient by performing secondary inverse-transformation on a low frequency component of the inverse-quantization coefficient, the secondary inverse-transformation corresponding to the secondary transformation; and perform core inverse-transformation on the secondary inverse-transformation coefficient, the core inverse-transformation corresponding to the core transformation.