Video Coding Block Segmentation for Occlusion Handling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding schemes face inefficiencies in handling occlusions and disocclusions, particularly along object boundaries, due to redundant motion information and high computational complexity in geometric motion partitioning, which affects coding efficiency and data rate.
Innovation Solution
The implementation of segmentation-based partitioning for inter prediction of video coding blocks, using motion-dependent filtering units to apply directional smoothing or feathering filters based on divergence measures of segment motion vectors, allowing for improved handling of occlusions and disocclusions by adapting filter properties and types.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If geometric motion partitioning (GMP) is used to partition video coding blocks along object boundaries, then coding efficiency is improved, but computational complexity increases greatly
Solution Approach 1:
The patent applies segmentation by dividing the video coding block into multiple segments based on detected object boundaries. Each segment is then processed with motion estimation independently, allowing the system to capture different motion characteristics in different regions without requiring exhaustive search across the entire block, thus reducing computational complexity while maintaining coding efficiency.
Solution Approach 2:
The patent implements local quality by applying different motion estimation strategies to different segments of the video block. Regions with object boundaries receive specialized handling with boundary-aware motion estimation, while uniform regions use standard motion estimation, optimizing the balance between coding efficiency and computational complexity locally rather than uniformly across the entire block.
2Measurement precision
If fine quadtree decomposition and block partitioning are applied along object boundaries, then motion information accuracy is improved, but redundancy is introduced decreasing coding efficiency
Solution Approach 1:
The patent applies dynamics by adaptively adjusting the block partitioning structure based on the detected object boundary characteristics. Rather than using fixed fine quadtree decomposition everywhere, the system dynamically selects partitioning depths and patterns according to local boundary complexity, ensuring high motion information accuracy at boundaries while avoiding unnecessary redundancy in simpler regions.
3Ease of manufacture
If block based inter prediction methods are used, then processing simplicity is maintained, but handling of occlusions and disocclusions is inconsistent
Solution Approach 1:
The patent segments the video block along detected object boundaries and applies specialized motion estimation to boundary segments. This segmentation allows the system to handle occlusions and disocclusions more reliably by recognizing boundary regions and applying appropriate prediction strategies, while maintaining simplicity in non-boundary regions through standard block-based prediction.
Solution Approach 2:
The patent applies local quality by using different prediction approaches for boundary regions versus interior regions. Boundary regions with potential occlusions receive enhanced handling with multiple reference frame consideration and boundary-aware motion estimation, while interior regions maintain simple block-based prediction, thus improving occlusion handling consistency without sacrificing overall processing simplicity.
Data Source
Figure 1a
Figure 1b
Figure 2a
AI summary
The invention relates to a decoding apparatus for decoding a video coding block. The decoding apparatus comprises a decoding unit decoding the video coding block for providing a residual video coding block; a partitioning unit partitioning the video coding block on the basis of the coding information into two or more segments including a first segment and a second segment. The coding information comprises a first segment motion vector associated with the first segment of the video coding block and a second segment motion vector associated with the second segment of the video coding block. An inter prediction unit determines on the basis of the first segment motion vector a co-located first segment in a first reference frame and on the basis of the second segment motion vector a co-located second segment in a second reference frame and generates a predicted video coding block on the basis of the co-located first segment and the co-located second segment. The predicted video coding block comprises a predicted first segment and a predicted second segment. A motion-dependent filtering unit determines a divergence measure on the basis of the first segment motion vector and the second segment motion vector and applies depending on the divergence measure a first filter or a second filter to the predicted video coding block. A reconstruction unit reconstructs the video coding block on the basis of the filtered predicted video coding block and the residual video coding block.