In-loop and Post-filtering for Image Boundary Distortion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image encoding methods cause distortion at processing boundaries when areas divided as tiles or slices are processed in parallel, requiring additional processes and memory for filtering.
Innovation Solution
An image decoding method that filters boundaries between reconstructed blocks and decoded neighboring blocks using different filter strengths, with switching based on boundary conditions, allowing in-loop and post-filtering, and recording filter information in the bitstream for encoder-side switching.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If boundaries are filtered to suppress distortion, then image quality is improved, but processing complexity and memory requirements increase
Solution Approach 1:
The patent divides the picture into multiple tiles that can be processed in parallel. By segmenting the image processing into independent tile units, the system can apply filtering operations to individual tiles without requiring complex coordination across the entire image, thus reducing overall processing complexity while maintaining boundary distortion suppression.
Solution Approach 2:
The patent applies filtering operations during the encoding process before the actual parallel processing of tiles. By performing preliminary filtering on boundary blocks during encoding, the system prepares the data in advance so that subsequent parallel decoding and processing operations can proceed without requiring additional complex filtering, thereby reducing processing complexity during the main computation phase.
2Manufacturing precision
If filtering is applied to suppress boundary distortion, then image quality is improved, but processing time increases
Solution Approach 1:
The patent performs filtering operations during the encoding phase rather than during decoding or post-processing. By applying the necessary filtering in advance during encoding, the system eliminates the need for time-consuming filtering operations during the critical parallel processing phase, thus reducing overall processing time while maintaining image quality.
Solution Approach 2:
The patent segments the image into tiles that can be processed independently and in parallel. This segmentation allows the system to process multiple tiles simultaneously without requiring sequential filtering operations across the entire image, significantly reducing the total processing time while still applying necessary filtering to suppress boundary distortions at tile edges.
3Productivity
If parallel processing is used to increase encoding efficiency, then productivity is improved, but boundary distortion occurs
Solution Approach 1:
The patent divides the image into multiple independent tiles that can be processed in parallel by different processing units. This segmentation enables simultaneous processing of multiple regions without interference, maintaining high encoding efficiency while allowing each tile to be processed independently to avoid boundary distortion issues that would arise from processing the entire image as a single unit.
Solution Approach 2:
The patent applies filtering operations to boundary blocks during the encoding phase before parallel processing begins. By performing this preliminary filtering action on blocks that will form boundaries between tiles, the system ensures that when parallel processing occurs, the boundary regions are already prepared and will not exhibit distortion, thus maintaining both productivity and image quality.
Data Source
AI summary
A decoding method includes: predicting a current block in an image using a reference block, to generate a prediction block; and generating a reconstructed block using the prediction block, wherein the generating includes: filtering for a boundary between the reconstructed block and a decoded neighboring block, using a first filter strength set using first prediction information for prediction of the current block and second prediction information for prediction of the decoded neighboring block; filtering for the boundary using a second filter strength set without using the second prediction information of the current block and the decoded neighboring block; and switching whether to execute the second filtering, based on the boundary, wherein the first filtering is in-loop filtering in which a filtered reconstructed block is used as a reference block for another block, and the second filtering is post filtering outside the loop.