Content-Adaptive Binocular Matching Using Grid Reference Points
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Binocular stereo matching in three-dimensional perception is hindered by high computational complexity and accuracy issues, particularly in variable environments, due to noise, occlusions, and non-ideal factors, limiting its real-time performance and practical applications.
Innovation Solution
A content-adaptive binocular matching method that divides images into grids to calculate feature information, determines reference points with high matching degrees, and uses depth-direction layering to establish parallax values and search ranges, ensuring accurate matching while reducing computational complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If global algorithm is used for binocular matching, then matching accuracy is improved, but computational complexity increases and real-time performance deteriorates
Solution Approach 1:
The patent divides the image into multiple grids and performs matching calculations at grid points to determine reference points. This segmentation approach reduces the overall computational complexity by breaking down the global matching problem into smaller local problems, while still maintaining matching accuracy through the use of reference points for guiding subsequent matching.
Solution Approach 2:
The patent implements depth-direction layering that divides the image into different depth layers, where each layer has adaptive search ranges. This local quality approach allows different regions to have different matching parameters optimized for their specific characteristics, improving matching accuracy in complex regions while reducing computational burden in simpler regions.
2Measurement precision
If global algorithm is used for binocular matching, then matching accuracy is improved, but processing speed decreases
Solution Approach 1:
The patent performs preliminary matching calculations at grid points to determine reference points before conducting full image matching. This preliminary action identifies key reference points that guide subsequent matching operations, reducing the overall processing time while maintaining accuracy by using these pre-identified references to constrain the search space.
Solution Approach 2:
The patent implements dynamic adaptive search ranges based on depth layers and reference points. The search range adjusts dynamically according to the depth layer and local image characteristics, allowing faster processing in regions with clear features while maintaining accuracy in complex regions through expanded search ranges.
3Measurement precision
If large search range is used in local algorithm, then matching accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent changes the search range parameter dynamically based on depth layers and reference points. Instead of using a fixed large search range throughout the image, the search range is adaptively adjusted according to local image characteristics and depth information, reducing computational complexity in regions where smaller ranges suffice while maintaining accuracy where larger ranges are needed.
4Productivity
If small search range is used in local algorithm, then computational complexity is reduced, but matching accuracy deteriorates
Solution Approach 1:
The patent applies different search range sizes to different local regions based on their characteristics. Regions with simple features use smaller search ranges for faster processing, while regions with complex features or occlusions use larger search ranges to maintain matching accuracy, achieving a balance between speed and precision across the entire image.
Data Source
AI summary
Provided are a content-adaptive binocular matching method and apparatus. The method includes: dividing both left and right images into grids so as to calculate feature information at grid points, and performing matching calculation between the left and right images at the grid points according to the feature information so as to determine points with high matching degrees as reference points; calculating parallax values of the reference points, determining an interlayer spacing, and layering all the reference points in a depth direction according to the parallax values and the interlayer spacing; for a non-reference point pixel in a certain image, determining a reference point closest to the non-reference point pixel, and determining a search range and the parallax value of the reference point; and performing matching calculation on the non-reference point pixel and a pixel in the search range in the other image.