Video Decoder Deblocking Filter for Asymmetric Block Boundaries
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Symmetrical deblocking filters in existing video coding standards like H.265/HEVC are inefficient in reducing pixel value errors when there is a discontinuity in error distribution across block boundaries, leading to increased errors in pixel values.
Innovation Solution
Implementing asymmetrical deblocking filters across block boundaries by adjusting filter characteristics, such as coefficients, weights, offsets, or conditions, based on pixel values, block sizes, prediction directions, and quantization parameters to selectively apply more filtering to pixels with larger errors and less to those with smaller errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If symmetrical deblocking filters are used across block boundaries, then the filtering process is simple and consistent, but pixel value errors are not effectively reduced when error distribution is discontinuous
Solution Approach 1:
The patent applies asymmetry by using different filter coefficients for pixels on different sides of a block boundary. Specifically, when a boundary is detected between blocks with different prediction types or modes, the filter coefficients are made asymmetric: one side uses standard coefficients while the other side uses modified coefficients (e.g., doubled coefficients) to better match the discontinuous error distribution characteristics.
Solution Approach 2:
The patent implements local quality by adapting filter characteristics to local regions around block boundaries. Different filter coefficients are applied to different locations relative to the boundary based on detected characteristics such as prediction type changes, mode differences, or boundary strength metrics, allowing each local region to receive optimized filtering.
2Manufacturing precision
If filter coefficients are adjusted to handle discontinuous error distribution, then pixel value errors are reduced, but filter design complexity increases
Solution Approach 1:
The patent applies dynamics by making filter coefficients adaptive rather than fixed. The filter characteristics are dynamically adjusted based on detected boundary conditions, prediction types, and error distribution patterns. This allows the system to automatically select appropriate filter coefficients for different scenarios without requiring complex manual design.
Solution Approach 2:
The patent implements parameter changes by modifying filter coefficients based on detected characteristics such as prediction type changes, mode differences, or boundary strength. The coefficients are scaled or transformed according to these parameters (e.g., doubling coefficients when certain conditions are detected), allowing a single filter structure to handle multiple error distribution patterns.
Data Source
AI summary
Various embodiments provide a decoder configured to select a filter based on a block size of a first block and a block size of a second block in an image, and change values of pixels in the first block and the second block. The filter includes a first set of multipliers and a first set of offsets for the first block, and a second set of multipliers and a second set of offsets for the second block. The values of the pixels in the first block and the second block are changed by performing multiplication with each multiplier in the first set of multipliers, by performing multiplication with each multiplier in the second set of multipliers, and by using the first set of offsets and the second set of offsets.