Visual Data Codeword Segmentation for Efficient Entropy Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image and video compression technologies, both classical and neural-network-based, face challenges in achieving optimal coding efficiency and effectiveness, particularly in modeling probability distributions and handling spatial dependencies in latent variables, which limits their performance in lossy compression scenarios.
Innovation Solution
The proposed method involves using a plurality of codewords associated with semantic element information or latent variable information of visual data units, incorporating a hyperprior model and autoregressive components to enhance entropy coding, thereby improving coding efficiency and effectiveness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single codeword is used for compression, then the device complexity is reduced, but the coding efficiency deteriorates
Solution Approach 1:
The patent divides the compression data structure into multiple codewords, where each codeword represents a specific color component (Y, U, V) or semantic element. This segmentation allows independent optimization of each codeword's probability distribution modeling, improving coding efficiency without significantly increasing overall system complexity.
2Reliability
If classical transform-based methods are used, then the coding effectiveness is maintained, but the ability to model spatial dependencies in latent variables deteriorates
Solution Approach 1:
The patent introduces a neural network as an intermediary component that processes visual data before entropy coding. This neural network learns and models complex spatial dependencies in the data, transforming the input into a representation that can be more effectively compressed by traditional entropy coding methods, thus combining the strengths of both approaches.
3Adaptability or versatility
If neural network-based compression is used, then the spatial dependency modeling is improved, but the device complexity increases
Solution Approach 1:
The patent segments the neural network processing into separate branches for different color components (luma and chroma). Each branch independently models spatial dependencies for its specific component, allowing the system to benefit from neural network capabilities while reducing the overall computational complexity compared to a monolithic neural network approach.
Data Source
AI summary
Embodiments of the present disclosure provide a solution for visual data processing. In the method, for a conversion between a current visual unit of visual data and a bitstream of the visual data, a plurality of codewords in the bitstream is determined. A codeword is associated with at least one of: semantic element information, or latent variable information of at least one color component of the current visual unit. The conversion is performed based on the plurality of codewords.