Deblocking Filter Memory Management for HEVC
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing image processing systems for HEVC require a large number of line memories due to the sequential processing of deblocking filters and sample adaptive offset filters, leading to frequent access to external memory, which is inefficient.
Innovation Solution
The proposed solution involves an image processing apparatus with a deblocking filter and a sample adaptive offset processing unit that stores and reads blocks from memory only when necessary, reducing external memory access by performing filter processes in a pipeline manner and using intermediate memories to store pixel data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If filtering is performed for the entire vertical edge on a picture basis and then for the entire horizontal edge, then image quality is improved by removing block noise, but the number of line memories required increases greatly
Solution Approach 1:
The patent divides the picture into multiple blocks and processes each block independently through pipeline stages. Instead of holding entire picture edges in memory, only current and previous blocks are maintained in pipeline buffers, reducing memory requirements while maintaining filtering quality
Solution Approach 2:
The patent performs preliminary decoding and prepares block data in advance before filtering is needed. By pre-decoding blocks and maintaining them in pipeline buffers, the system avoids frequent external memory accesses during the filtering stage
2Productivity
If a deblocking filter and sample adaptive offset filter are applied sequentially on an LCU basis, then coding efficiency is improved, but the number of external memory accesses increases
Solution Approach 1:
The patent implements a continuous pipeline that processes multiple blocks simultaneously through overlapping stages of deblocking filtering and sample adaptive offset filtering. This eliminates idle time between operations and maintains continuous data flow, preventing external memory accesses
Solution Approach 2:
The system pre-decodes and pre-processes blocks before they reach the filtering stages, so that when filtering occurs, all necessary data is already available in the pipeline buffers, eliminating the need for external memory accesses during filtering operations
3Measurement precision
If filtering is performed on an LCU basis with line memories for each LCU edge, then local decoding accuracy is improved, but hardware complexity increases
Solution Approach 1:
The patent segments the filtering process into discrete pipeline stages that handle individual blocks rather than requiring separate line memories for each LCU edge. This modular approach maintains processing accuracy while reducing hardware complexity
Solution Approach 2:
The pipeline buffers serve multiple functions: they store decoded blocks, provide data for deblocking filtering, and supply input for sample adaptive offset filtering. This multi-functionality eliminates the need for separate dedicated line memories for each processing stage
Data Source
AI summary
An image processing apparatus includes a decoding unit, a deblocking filter, and a sample adaptive offset processing unit. The decoding unit is configured to perform a decoding process for a coded stream to generate an image. The deblocking filter is configured to store a block, a corresponding block of which on a current edge is not input, in a memory, out of blocks of the image generated by the decoding unit and read the block stored in the memory at timing when the corresponding block is input, to perform a filter process. The sample adaptive offset processing unit is configured to perform a sample adaptive offset process by using a pixel of the block that has been subjected to the filter process by the deblocking filter.