In-loop Deblocking Filter for Large Coding Unit Artifact Removal
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional HEVC deblocking filters fail to effectively smooth discontinuities at the edges of large coding units (CUs), leading to block artifacts despite the application of strong filtering conditions.
Innovation Solution
An improved in-loop deblocking filter apparatus that processes current rows or columns of samples by applying a (2N-1) tap filter with specific coefficients, determining filtered sample values based on conditions that assess the differences between sample values across edges.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If the HEVC strong deblocking filter is applied to large CUs, then the filtering strength is increased, but block artifacts persist at the shorter edges and large CU boundaries
Solution Approach 1:
The patent divides the filtering process into two distinct stages: a first deblocking filter for normal edges and a second deblocking filter for edges meeting specific criteria (large CUs with width/height > 8 samples). This segmentation allows each filter to be optimized for its specific purpose, with the second filter applying stronger smoothing only where needed, thereby removing block artifacts from large CU boundaries without unnecessarily processing all edges.
Solution Approach 2:
The patent applies different filtering strengths to different locations based on local characteristics. The second deblocking filter with stronger smoothing effect is applied selectively only to edges of large CUs (width/height > 8 samples), while normal edges receive the standard filter strength. This local quality approach ensures that block artifacts are addressed where they occur most prominently without degrading the quality of other regions.
2Manufacturing precision
If a longer tap filter is applied to smooth discontinuities at large CU edges, then the smoothing effect is improved, but the computational complexity increases
Solution Approach 1:
The patent segments the filtering operation into two passes: a first deblocking filter that processes all edges with standard complexity, and a second deblocking filter that processes only edges meeting specific criteria (large CUs). This segmentation reduces the overall computational complexity by limiting the application of the more complex second filter to only those edges where it is most needed, rather than applying it universally.
Solution Approach 2:
The patent applies the second deblocking filter with stronger smoothing effect partially, only to edges where block artifacts are most prominent (large CUs with width/height > 8 samples). This partial action approach avoids the excessive computational cost of applying the same strong filtering to all edges, while still achieving effective artifact removal where it matters most.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An in-loop deblocking filter apparatus (120) for processing a current row or column of samples into a filtered row or column of samples. The current row or column of samples comprises a plurality of samples from a first sample block and a horizontally or vertically neighboring second sample block of a reconstructed picture of a video stream. The samples of the current row or column of samples have sample values p N-1 -p 0, q 0 -q N-1wherein N is an even integer greater than 2. If a first condition or a second condition is satisfied, the current row or column is processed by: determining a filtered sample value q' 0 by applying a (2N-1) -tap filter to the sample values p N-2, …, p 0, q 0, …, q N-1 of the current row or column; and/or determining a filtered sample value p' 0 by applying a (2N-1) -tap filter to the sample values p N-1, …, p 0, q 0, …, q N-2 of the current row or column.