Shared Context Memory for Parallel Image Coding Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The wavefront method for image coding, which involves parallel processing of context-based adaptive binary arithmetic coding (CABAC), leads to significant deterioration in coding efficiency due to the need for context parameter initialization and copying between block lines, resulting in processing stops during parameter copy operations.
Innovation Solution
An image coding apparatus that eliminates the need for context parameter copying by using a shared context memory to store updated parameters, allowing each entropy coding unit to encode and decode block lines based on the shared context, thereby avoiding the need for parameter initialization and copying between block lines.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If parallel processing is performed for each block line using the wavefront method, then processing time is reduced, but coding efficiency deteriorates significantly due to context parameter initialization and copying
Solution Approach 1:
The picture is divided into multiple block lines that can be processed in parallel. Each entropy coding unit processes a specific block line independently, allowing simultaneous processing of multiple block lines while maintaining context parameter continuity through the shared context memory structure.
Solution Approach 2:
A shared context memory is introduced as an intermediary between entropy coding units processing different block lines. This shared memory stores context parameters that can be accessed by multiple entropy coding units, eliminating the need for parameter copying and initialization while enabling parallel processing.
2Productivity
If context parameter initialization is performed before coding processing for each block line, then parallel processing is enabled, but coding efficiency deteriorates
Solution Approach 1:
Context parameters are prepared in advance in the shared context memory before parallel processing begins. This preliminary preparation allows multiple entropy coding units to access the same context parameters simultaneously without requiring initialization during parallel processing, thus maintaining coding efficiency while enabling parallel operation.
Solution Approach 2:
The shared context memory serves multiple functions: it stores context parameters for current processing, maintains parameters for parallel processing units, and eliminates the need for repeated initialization. This multi-functional design allows the same memory structure to support both parallel processing and parameter continuity.
3Productivity
If context parameter copying is performed between block lines, then parallel processing is maintained, but processing stops during copy operations
Solution Approach 1:
Multiple context memory structures are merged into a single shared context memory that is accessible by all entropy coding units. This consolidation eliminates the need for parameter copying between separate memory structures, as all units access the same shared memory, thus removing processing stops while maintaining parallel processing continuity.
Solution Approach 2:
Instead of physically copying context parameters between memory structures, the system uses shared memory addressing where multiple entropy coding units reference the same context parameters in the shared context memory. This virtual copying approach maintains parameter consistency without actual data transfer, eliminating processing delays.
Data Source
AI summary
To prevent deterioration in performance that is caused by the processing for copying a context parameter between memories, an image processing apparatus writes an updated context parameter into context memories corresponding to respective entropy coding units and into a shared context memory, until it completes encoding processing for M blocks positioned on the left side of a slice.