Sample Adaptive Offset Band Processing for Faster HEVC BO Mode
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing Band Offset (BO) mode in HEVC's sample adaptive offset (SAO) technology is time-consuming and increases power consumption, particularly in scenarios like live streaming, due to the extensive calculations of residual values and pixel counts in coding tree blocks.
Innovation Solution
A method is introduced where the band values and residual values of reconstructed pixels are pre-calculated and stored in arrays, allowing for cumulative calculations of consecutive band values belonging to the same band, reducing memory read/write operations and accelerating data processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the existing BO mode performs extensive calculations of residual values and pixel counts in coding tree blocks, then the pixel compensation accuracy is improved, but the processing time and power consumption increase significantly
Solution Approach 1:
The patent pre-calculates and stores band values and residual values of reconstructed pixels in arrays before the main processing stage. This preliminary action eliminates the need for repeated calculations during the actual BO mode processing, significantly reducing processing time while maintaining pixel compensation accuracy.
Solution Approach 2:
The patent divides the image processing into distinct segments: first calculating and storing band values in a first array, then calculating and storing residual values in a second array. This segmentation allows each calculation to be performed once and reused, reducing overall processing time and power consumption.
2Measurement precision
If the existing BO mode performs extensive calculations of residual values and pixel counts, then the pixel compensation accuracy is improved, but the power consumption increases
Solution Approach 1:
The patent performs band value and residual value calculations in advance and stores them in arrays. This preliminary computation eliminates redundant calculations during main processing, significantly reducing power consumption while preserving the accuracy required for effective pixel compensation in BO mode.
Solution Approach 2:
The patent creates copies of calculated values (band values and residual values) and stores them in arrays for repeated use. Instead of recalculating these values multiple times, the system copies and reuses the pre-computed data, reducing energy consumption while maintaining compensation accuracy.
3Productivity
If the patent pre-calculates and stores band values and residual values in arrays, then the processing speed is improved, but the memory usage increases
Solution Approach 1:
The patent pre-calculates band values and residual values and stores them in arrays to enable faster processing during BO mode. This trade-off increases memory usage but significantly improves processing speed by eliminating redundant calculations.
Solution Approach 2:
The patent extracts only the essential values (band values and residual values) that are needed for BO mode processing and stores them in arrays. This selective extraction minimizes memory usage while still providing the speed benefits of pre-computation.
Data Source
AI summary
A method is provided that includes: traversing all reconstructed pixels in a coding tree block, calculating a band value of each traversed reconstructed pixel, and calculating a residual value between each traversed reconstructed pixel and a corresponding original pixel; storing the calculated band value of each reconstructed pixel in a first array, and storing the calculated residual value between each reconstructed pixel and the corresponding original pixel in a second array; determining whether m consecutive band values in the first array belong to a same band; and in response to determining that the m consecutive band values in the first array belong to the same band, calculating, based on a residual value in the second array, a cumulative residual value of m reconstructed pixels corresponding to the m consecutive band values, and calculating a cumulative number of the m reconstructed pixels.