Image Coding With Reinitialized Entropy Coding for Parallel Decoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding methods, such as MPEG and H.264, suffer from inefficiencies due to sequential entropy coding that leads to loss of local probability conformity and restricts parallel decoding, especially in multi-core architectures, and do not effectively utilize local symbol probabilities for improved compression.
Innovation Solution
The method involves segmenting images into subsets of blocks, using previously determined symbol occurrence probabilities from other subsets for the first block of each subset, and implementing a reinitialization of entropy coding modules to allow parallel decoding and coding, independent of the coding order.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If sequential entropy coding is used for video compression, then compression efficiency is improved through probability learning, but parallel decoding is restricted and processing time increases
Solution Approach 1:
The image is divided into multiple subsets of blocks, where each subset can be independently decoded. This segmentation enables parallel decoding of different subsets while maintaining compression efficiency through context-adaptive probability learning within each subset.
Solution Approach 2:
Symbol occurrence probabilities are determined preliminarily for each subset based on previously decoded subsets, allowing the entropy coding module to be reinitialized with accurate probability information before decoding each subset, thus enabling both efficient compression and parallel processing.
2Loss of energy
If sequential entropy coding with probability learning is implemented, then local symbol probabilities are effectively utilized for compression, but buffer memory requirements increase
Solution Approach 1:
By segmenting the image into subsets that can be independently decoded, the buffer memory requirements are reduced since only the current and previously decoded subsets need to be stored, rather than maintaining large buffers for sequential processing of the entire image.
Solution Approach 2:
The entropy coding module maintains separate probability information for each subset, allowing local adaptation to the statistical characteristics of each subset while reducing the overall memory burden through localized probability learning rather than global probability tracking.
3Loss of energy
If context-adaptive probability learning is applied across the entire image, then compression efficiency is maximized, but parallel decoding becomes difficult
Solution Approach 1:
The image is segmented into independent subsets that can be decoded in parallel, with each subset undergoing context-adaptive probability learning locally. This maintains compression efficiency while enabling parallel processing capabilities.
Solution Approach 2:
The entropy coding module is designed to be reinitializable with probability information from previously decoded subsets, allowing it to function both in sequential mode for maximum compression and in parallel mode for improved processing speed, thus providing multi-functional adaptability.
Data Source
AI summary
A method for coding includes; segmenting an image into blocks; grouping blocks into a number of subsets; coding, using an entropy coding module, each subset, by associating digital information with symbols of each block of a subset, including, for the first block of the image, initializing state variables of the coding module; and generating a data sub-stream representative of at least one of the coded subsets of blocks. Where a current block is the first block to be coded of a subset, symbol occurrence probabilities for the first current block are determined based on those for a coded and decoded predetermined block of at least one other subset. Where the current block is the last coded block of the subset: writing, in the sub-stream representative of the subset, the entire the digital information associated with the symbols during coding of the blocks of the subset, and implementing the initializing sub-step.