Intra Prediction Mode Selection Using Texture Gradient Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The complexity of choosing optimal intra prediction modes and motion estimation in video encoding, particularly in High Efficiency Video Coding (HEVC), leads to high computational complexity due to a large number of possible prediction angles and block sizes, which hinders efficient video compression.
Innovation Solution
The proposed methods, including Algorithms 1-9, reduce the complexity by selecting optimal intra prediction modes using Minimal Activity Directions and efficient motion estimation techniques for inter coding, such as multi-pass motion vector refinement and local transform-based estimation, to streamline the video encoding process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If exhaustive search of all possible intra prediction angles is performed, then optimal prediction accuracy is improved, but computational complexity increases significantly
Solution Approach 1:
The patent segments the exhaustive search space of 33 intra prediction angles into multiple subsets or groups. Instead of evaluating all angles uniformly, the method divides them into categories (e.g., vertical, horizontal, diagonal groups) and selectively searches based on block characteristics, reducing the number of angles that need to be evaluated while maintaining prediction accuracy.
Solution Approach 2:
The patent applies partial action by evaluating only a subset of prediction angles rather than all possible angles. Based on block texture analysis and gradient calculations, the method identifies and evaluates only the most relevant angles for each block, performing partial searches that suffice for achieving optimal prediction without the need for exhaustive evaluation of all 33 angles.
2Adaptability or versatility
If a large number of intra prediction angles are supported, then prediction flexibility is improved, but mode selection complexity increases
Solution Approach 1:
The patent implements dynamic mode selection where the number and types of prediction angles evaluated are adapted based on block characteristics. The method dynamically adjusts the prediction angle search strategy according to block size, texture complexity, and gradient directions, evaluating fewer angles for simple blocks and more angles for complex blocks, thereby maintaining flexibility while reducing average complexity.
Solution Approach 2:
The patent applies local quality by using different prediction angle evaluation strategies for different regions or types of blocks. Based on local texture analysis and gradient calculations, the method selects appropriate prediction angles tailored to each block's local characteristics, rather than applying a uniform exhaustive search to all blocks, thus maintaining adaptability while reducing overall complexity.
3Measurement precision
If motion estimation is performed for all possible block sizes and partitions, then motion compensation accuracy is improved, but encoding time increases
Solution Approach 1:
The patent performs preliminary actions by conducting coarse motion estimation at reduced resolution or with limited block sizes before performing fine motion estimation. The method initially evaluates a smaller set of candidate motion vectors and block partitions, then refines the search only for promising candidates, avoiding exhaustive motion estimation for all possible block sizes and partitions from the outset.
Solution Approach 2:
The patent applies skipping by bypassing exhaustive motion estimation for block sizes and partitions that are unlikely to provide significant improvement. Based on preliminary analysis and gradient-based predictions, the method skips detailed motion estimation for certain block configurations and directly uses simplified motion compensation, thus reducing encoding time while maintaining accuracy for critical blocks.
Data Source
AI summary
Simplified selection of optimal intra prediction in block-based video coding based on texture gradient distribution and minimal activity direction. The minimal activity direction is defined by a vector (α(B, W), β(B, W)) indicating minimal variation of a discrete function P(x, y) inside a spatial aria of block B with weights W.