Stereo Matching Search Range Optimization via Reference Disparity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing stereo matching processing methods are inefficient and prone to mismatching errors due to the need to search along the entire epipolar line, which increases processing time and error rates when generating three-dimensional data from stereo images.
Innovation Solution
A stereo matching processing apparatus and method that sets a reference disparity and a smaller search range based on the difference in height and estimated height of terrestrial objects, allowing for faster and more accurate matching by restricting the search range to areas where matching points are likely to exist.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the search window is restricted to move only in the direction of an epipolar line, then the amount of processing is reduced, but the processing time increases when searching all along the epipolar line
Solution Approach 1:
The patent divides the epipolar line search into multiple discrete sampling points spaced at predetermined intervals rather than searching continuously along the entire epipolar line. This segmentation reduces the number of search positions from potentially thousands to a manageable set of sampled points, thereby reducing processing time while maintaining acceptable matching accuracy.
Solution Approach 2:
The patent performs partial search action by sampling only at specific intervals along the epipolar line rather than exhaustively searching every possible position. This partial action approach achieves a balance between search thoroughness and processing speed, obtaining sufficient matching information without the computational burden of complete enumeration.
2Ease of operation
If the search window is restricted to move only on the scanning line after coordinate transformation, then the disparity can be obtained as a difference between x-coordinate values, but mismatching probabilities increase
Solution Approach 1:
The patent performs preliminary coordinate transformation to align the epipolar line with the scanning line direction before conducting the matching search. This preliminary action simplifies the subsequent search process by allowing the search window to move only along the scanning line, making disparity calculation straightforward as the difference between x-coordinate values of matching points in left and right images.
Solution Approach 2:
The patent combines coordinate transformation with segmented sampling by dividing the transformed epipolar line into discrete sampling points at predetermined intervals. This combination maintains the computational simplicity of scanning line search while reducing mismatching probabilities through strategic sampling at multiple positions rather than relying on a single search point.
3Reliability
If the entire epipolar line is searched for matching points, then comprehensive matching is achieved, but the processing time becomes excessively long
Solution Approach 1:
The patent segments the continuous epipolar line into discrete sampling points spaced at predetermined intervals. Instead of searching every possible position along the epipolar line, the search is performed only at these sampled points. This segmentation maintains matching completeness by covering key positions along the epipolar line while dramatically reducing the total number of search operations required.
Solution Approach 2:
The patent applies partial action by performing searches at selected intervals rather than exhaustively searching the entire epipolar line. This approach obtains sufficient matching information from strategically positioned sample points, achieving acceptable matching completeness without the computational expense of complete enumeration of all possible positions.
Data Source
AI summary
Improving stereo matching speed and accuracy, an image data input unit acquires image data of plural images of a predetermined region captured from plural different positions. A reference disparity setting unit sets a reference disparity suitable for the plural images. The search range setting unit sets a predetermined range smaller than the image range as a search range for stereo matching, by referring to points in the images between which the reference disparity set by the reference disparity setting unit is provided. A stereo matching unit searches out, for an arbitrary point in one of the plural images, a point in the other image that matches the arbitrary point, from the search range set by the search range setting unit, by referring to a point in the other image that provides the reference disparity set by the reference disparity setting unit.