Pattern-Based Block Filtering for Gradient-Adaptive Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video encoding techniques using bilateral filtering (BLF) face limitations due to a fixed cross pattern that reduces the number of samples used for filtering, leading to decreased efficiency, particularly for diagonal gradients, and increased complexity.
Innovation Solution
Introduce pattern diversity in block filtering by selecting filter patterns based on block properties, such as local gradients, allowing for various shapes like cross, diagonal, and square patterns, and signaling these patterns in the bitstream or determining them at the decoder.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a fixed cross pattern is used for bilateral filtering, then the filtering process is simple, but the number of samples used for filtering is reduced, leading to decreased compression efficiency and poor performance for diagonal gradients
Solution Approach 1:
The patent applies dynamics by making the filter pattern adaptive rather than fixed. The filter pattern is dynamically selected based on the gradient direction calculated from the block, allowing the filtering process to adapt to different image content characteristics. This resolves the contradiction by enabling the system to use more appropriate samples for filtering (improving compression efficiency) while only adding minimal computational complexity for gradient calculation and pattern selection.
Solution Approach 2:
The patent changes the parameter of filter pattern selection from a fixed value to a variable determined by local image characteristics (gradient direction). By calculating gradients and selecting patterns based on these parameters, the system can utilize more relevant samples for filtering in different regions, thereby improving compression efficiency without substantially increasing overall complexity through standardized parameter calculation methods.
2Reliability
If more samples are used for filtering, then noise reduction improves, but the computational complexity increases
Solution Approach 1:
The patent applies local quality by selecting different filter patterns for different regions based on their local gradient characteristics. Instead of uniformly using more samples across the entire image (which would increase overall complexity), the system locally adapts the filtering approach to match the local image structure. This allows noise reduction to be improved in regions where it is beneficial while maintaining simpler processing in regions where it is not needed, thus resolving the contradiction between noise reduction quality and computational complexity.
3Manufacturing precision
If a fixed filtering approach is used, then the processing is fast, but it performs poorly for diagonal gradients and reduces filtering effectiveness
Solution Approach 1:
The patent makes the filtering approach dynamic by selecting patterns based on gradient direction rather than using a fixed approach. This dynamic adaptation allows the system to preserve edges more accurately by aligning the filtering direction with the local image structure, particularly for diagonal gradients. The minimal overhead of gradient calculation and pattern selection does not substantially impact processing speed, thus resolving the contradiction between edge preservation accuracy and filtering processing speed.
Data Source
AI summary
Methods and apparatuses for video coding and decoding are provided. The method of video encoding includes accessing a reconstructed block corresponding to a block in a picture of a video, determining at least one filter pattern based on a property of the block and filtering the reconstructed block according to the at least one filter pattern. The method of video decoding includes accessing a reconstructed block corresponding to a block in a picture of an encoded video, determining at least one filter pattern based on a property of the block and filtering the reconstructed block according to the at least one filter pattern. A bitstream formatted to include encoded data, a computer-readable storage medium and a computer program product are also described.