Stereo Matching Using Seed Point Disparity Propagation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current stereo matching algorithms face challenges in balancing speed and accuracy, with local algorithms being fast but ambiguous and global algorithms being accurate but non-real-time, and existing methods for disparity propagation being time-consuming.
Innovation Solution
A method that calculates initial disparity values using a local stereo matching algorithm, selects seed points based on a left-right consistency check, constructs confidence propagation regions based on color difference and spatial pixel-to-pixel distance, and determines disparity values for non-seed points to form a dense disparity map.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a global stereo matching algorithm is used to achieve better stereo matching effect, then accuracy is improved, but processing time increases and real-time effect cannot be achieved
Solution Approach 1:
The patent segments the image into super-pixels using region merging algorithms, dividing the complex global matching problem into smaller regional units. This allows selective processing of only certain regions (those containing seed points) rather than all pixels, significantly reducing computation time while maintaining accuracy in critical areas.
Solution Approach 2:
The patent applies different processing strategies to different regions: seed points undergo rigorous local matching with cost calculation and verification, while non-seed points use simpler propagation methods. This local quality approach ensures high accuracy where needed (at seed points and their neighborhoods) while saving time in other regions.
2Productivity
If a local stereo matching algorithm is used to achieve faster processing, then speed is improved, but stereo matching effect becomes ambiguous and accuracy decreases
Solution Approach 1:
The patent performs preliminary local stereo matching to identify and select seed points that satisfy specific conditions (cost difference thresholds, left-right consistency). These pre-selected seed points serve as reliable anchors for subsequent disparity propagation, ensuring accuracy is maintained while avoiding exhaustive global optimization.
Solution Approach 2:
The patent introduces seed points as intermediary elements that bridge local and global approaches. Seed points are computed using local matching but then serve as references for propagating disparity information across larger regions, effectively mediating between the speed of local methods and the accuracy of global methods.
3Productivity
If disparity propagation is performed using seed points from segmentation regions to balance speed and accuracy, then processing efficiency is improved, but early segmentation work becomes time-consuming
Solution Approach 1:
The patent implements dynamic region merging where super-pixels are adaptively formed based on local image characteristics and cost metrics. Regions are merged dynamically during the matching process rather than through static pre-segmentation, allowing the algorithm to focus computational resources on problematic areas while skipping well-behaved regions.
Solution Approach 2:
The patent performs partial segmentation only for regions containing seed points or regions of interest, rather than segmenting the entire image. This partial action approach reduces segmentation overhead while still providing the benefits of region-based processing where it matters most for accuracy.
Data Source
AI summary
Embodiments of the present invention relate to the field of machine vision technologies, and specifically disclose a method, apparatus and electronic device for stereo matching. By means of the foregoing technical solutions, the embodiments of the present invention can achieve an improvement in both a speed and an accuracy of stereo matching.