SAO Edge Offset Pixel Categorization With Single-Pass Block Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing methods for calculating pixel categories and residuals in the edge offset mode of sample adaptive offset (SAO) are time-consuming, necessitating the reading of reconstructed pixels multiple times, which significantly affects performance.
Innovation Solution
A method and apparatus for data processing in the edge offset mode of SAO that involves dividing a target reconstructed image into non-overlapping coding tree blocks, traversing pixels to calculate residual values and categories, and accumulating residual values and pixel counts based on predefined categories, reducing the number of times pixels need to be read.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If separate functions are written for each EO direction to complete pixel category calculation, then the calculation can be performed for all four directions, but the processing time increases significantly
Solution Approach 1:
The patent combines the calculation of pixel categories for all four edge offset directions into a single unified function. Instead of implementing separate functions for each direction (EO0, EO1, EO2, EO3), the invention merges them into one function that processes all directions simultaneously, thereby reducing processing time while maintaining the capability to handle all four directions.
Solution Approach 2:
The unified function serves multiple purposes by calculating pixel categories for all four edge offset directions within a single function call. This multi-functional approach eliminates the need for separate dedicated functions for each direction, reducing overall processing time while preserving adaptability to different edge orientations.
2Adaptability or versatility
If multiple separate functions are used for EO direction calculations, then each direction can be processed independently, but the number of function calls and reading operations increases
Solution Approach 1:
The patent merges multiple separate functions into a single unified function that handles all four edge offset directions. This consolidation reduces the number of function calls and reading operations required, while still maintaining the ability to process each direction independently through the unified function's internal logic.
3Measurement precision
If pixel compensation is performed using traditional EO mode with separate functions, then accurate pixel categorization is achieved, but computing time consumption increases
Solution Approach 1:
The patent combines multiple pixel categorization operations into a single unified function that processes all four edge offset directions simultaneously. This merging approach maintains the accuracy of pixel categorization while significantly improving computing speed by reducing the total number of operations required.
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
A data processing method and apparatus for an edge offset mode based on a sample adaptive offset. The method comprises: obtaining a target reconstructed image, and dividing the target reconstructed image into a plurality of non-overlapping coding tree blocks; for each coding tree block, traversing all reconstructed pixel points in the coding tree block, and calculating a residual value between each reconstructed pixel point and a corresponding original pixel point; traversing, in the coding tree blocks, all first reconstructed pixel points belonging to a non-boundary part and all second reconstructed pixel points belonging to a boundary part, and calculating categories of each traversed first reconstructed pixel point under edge offset modes in four directions and the category of each traversed second reconstructed pixel point under an edge offset mode in a single direction; and respectively calculating, according to the categories and the residual values, accumulated residual values of the reconstructed pixel points under different categories of the edge offset modes in four directions, and collecting statistics on the accumulated amount of the reconstructed pixel points under different categories.