Multi-Line Intra Interpolation Filters for Edge and Smooth Regions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies, such as HEVC and VVC, apply the same cubic and Gaussian interpolation filters to all reference lines in multi-line intra prediction, which is not desirable for areas with edges and smooth image regions.
Innovation Solution
A method and device for selecting different types of intra interpolation filters based on the reference line index, applying edge preserving filters to non-zero reference lines and edge smoothing filters to zero reference lines, with specific conditions and filter coefficients tailored for each line.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If the same cubic and Gaussian interpolation filters are applied to all reference lines in multi-line intra prediction, then the implementation is simple and uniform, but the prediction accuracy deteriorates for areas with edges and smooth image regions
Solution Approach 1:
The patent applies different interpolation filters to different reference lines based on their characteristics. Specifically, cubic filters are applied to reference lines that require edge preservation, while Gaussian filters are applied to reference lines that benefit from smoothing. This local differentiation of filter types resolves the contradiction by improving prediction accuracy for both edge and smooth regions without requiring a completely complex system, as the filter selection follows a systematic approach based on reference line properties.
2Manufacturing precision
If different types of intra interpolation filters are selected based on reference line index, then prediction accuracy and image quality are improved, but the computational complexity and processing time increase
Solution Approach 1:
The patent performs preliminary classification of reference lines by their indices before the actual interpolation process. Reference lines are pre-categorized into groups (e.g., first set of reference lines and second set of reference lines) based on their indices, and the appropriate filter type is predetermined for each group. This preliminary action allows the system to quickly select and apply the correct filter without performing complex real-time analysis during the interpolation process, thus improving image quality while minimizing additional processing time.
3Reliability
If multiple filter types are applied to different reference lines, then edge preservation and smoothing are optimized, but the device complexity and implementation difficulty increase
Solution Approach 1:
The patent segments the set of reference lines into distinct groups based on their indices. The first set of reference lines (with indices in a first range) is assigned cubic filters for edge preservation, while the second set of reference lines (with indices in a second range) is assigned Gaussian filters for smoothing. This segmentation simplifies the implementation by creating clear, manageable categories rather than requiring a complex, unified filter selection mechanism, thus optimizing edge preservation while controlling implementation complexity.
Data Source
AI summary
A method and apparatus for selecting an intra interpolation filter for multi-line intra prediction based on a reference line index for decoding a video sequence includes identifying a set of reference lines associated with a coding unit. A first type of interpolation filter is applied to reference samples included in a first reference line, of the set of reference lines, that is adjacent to the coding unit to generate a first set of prediction samples based on the first reference line being associated with a first reference line index. A second type of interpolation filter is applied to reference samples included in a second reference line, of the set of reference lines, that is non-adjacent to the coding unit to generate a second set of prediction samples based on the second reference line being associated with a second reference line index.