Video Coding Intra Prediction Mode Subset for Complexity Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing video coding schemes, such as those based on the AV1 standard, face substantial complexity when employing multiple reference lines for intra prediction, as they include all available intra prediction modes, leading to increased video coding complexity.
Innovation Solution
The proposed method limits the number of intra prediction modes for each non-zero reference line by defining an allowed intra prediction mode set (AIPM) that includes angular modes with nominal angles and either SMOOTH_V or SMOOTH_H modes, while excluding non-angular modes, thereby reducing video coding complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all available intra prediction modes are included for multiple reference lines, then prediction accuracy is improved, but video coding complexity increases substantially
Solution Approach 1:
The patent segments the set of intra prediction modes into different categories (angular modes with nominal angles, smooth modes, and non-angular modes) and applies different reference line configurations to each category. Specifically, angular modes are allowed on all reference lines while non-angular modes are restricted to only the zero reference line, creating a structured division that manages complexity while preserving accuracy where needed
Solution Approach 2:
The patent applies different quality levels of mode availability to different reference lines based on their importance. The zero reference line (immediately adjacent to the current block) receives the full set of intra prediction modes including non-angular modes, while non-zero reference lines receive a restricted set with only angular modes. This local differentiation maintains prediction accuracy for critical references while reducing complexity for less critical ones
2Adaptability or versatility
If multiple reference lines are applied with all intra prediction modes, then coding flexibility is improved, but the number of bits required to signal modes increases
Solution Approach 1:
The patent applies partial action by not including all intra prediction modes on all reference lines. Instead of providing the complete mode set uniformly, it selectively applies angular modes (which are more predictive for most content) on non-zero reference lines and reserves non-angular modes for only the zero reference line, achieving sufficient coding flexibility without the excessive bit cost of full mode availability everywhere
3Measurement precision
If non-angular intra prediction modes are included for non-zero reference lines, then prediction accuracy is improved, but video coding complexity increases
Solution Approach 1:
The patent performs preliminary action by pre-determining which intra prediction modes are appropriate for each reference line type before the actual prediction process. Non-angular modes are pre-assigned only to the zero reference line, while angular modes are pre-assigned to all reference lines. This preliminary classification eliminates the need for complex real-time decisions about mode applicability, reducing video coding complexity while maintaining prediction accuracy where it matters most
Data Source
AI summary
An example video decoding method includes obtaining a video bitstream that includes a plurality of encoded blocks including a current block. When a multiple reference line intra prediction (MRLP) mode is enabled for the current block and a non-adjacent reference line is used for the current block, a first allowed intra prediction mode (AIPM) set is determined for the current block. The first AIPM includes a set of one or more intra prediction modes of neighboring blocks of the current block, The first AIPM also includes a set of one or more intra prediction modes derived based on a signaled nominal angular mode and an angle delta value. The method further includes reconstructing the current block using an intra prediction mode from the first AIPM set.