Depth-Based Disparity Vector Determination for Video Boundary Areas
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video codecs face inefficiencies in encoding and decoding high-resolution video content, particularly when dealing with areas near the boundary of reference layer depth images, as they struggle to accurately determine disparity vectors, leading to suboptimal coding performance.
Innovation Solution
The method involves reconstructing color and depth images from a bitstream, determining depth-based disparity vectors for areas near the boundary of reference layer depth images, and using these vectors to directly or indirectly code current layer blocks, thereby improving coding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a video codec uses a limited encoding method based on coding units of a tree structure, then the encoding process can be simplified, but coding efficiency for high-resolution video content deteriorates
Solution Approach 1:
The patent divides the depth image into multiple depth image blocks corresponding to different depth ranges. Each block is processed independently to generate disparity vectors, allowing the system to handle high-resolution video content more efficiently by segmenting the complex task into manageable units while maintaining coding efficiency.
2Device complexity
If disparity vectors are determined without considering boundary areas of reference layer depth images, then the processing can be simplified, but prediction accuracy deteriorates
Solution Approach 1:
The patent applies different processing methods to different regions of the depth image. Specifically, it identifies boundary areas where depth image blocks deviate from the reference layer depth image boundary and applies special handling to these local regions. This ensures that prediction accuracy is maintained in critical boundary areas while simplifying processing in uniform regions.
Solution Approach 2:
The patent performs preliminary identification and classification of depth image blocks based on their spatial relationship with the reference layer depth image boundary. By pre-determining which blocks are boundary blocks and which are interior blocks, the system can apply appropriate processing methods in advance, improving prediction accuracy for boundary areas without increasing overall processing complexity.
3Device complexity
If depth values are not determined for areas deviating from the boundary of reference layer depth images, then the encoding process can be simplified, but coding performance deteriorates
Solution Approach 1:
The patent changes the parameter handling approach for depth values in boundary areas. Instead of using standard depth value determination methods for all regions, it applies modified parameter handling specifically for depth image blocks that deviate from the reference layer boundary. This allows the system to maintain coding performance in challenging boundary regions while keeping the overall encoding process simple.
Data Source
AI summary
An inter-layer decoding method including reconstructing a color image and a depth image of a first layer based on encoding information obtained from a bitstream; determining, from the depth image of the first layer, a depth image block of the first layer corresponding to a current block of a second layer image to be decoded; determining whether an area included in the determined depth image block deviates from a boundary of the depth image of the first layer; when the area included in the depth image block deviates from the boundary, determining a depth value of the area deviating from the boundary of the depth image; determining a disparity vector indicating a corresponding block of the first layer image with respect to the current block, based on the determined depth value; and reconstructing the current block by using the disparity vector.