Stereo Image Matching Using Dominant Orientation Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing stereo matching algorithms face challenges in efficiently generating accurate disparity maps, particularly in hardware implementation, due to high computational demands and error rates in texture-less regions or repetitive structures.
Innovation Solution
A method and device that determine dominant orientations in pixel locations of stereo images, detect points based on these orientations, and perform matching to estimate a sparse disparity map, filtering out non-informative pixels to reduce computational load and error rates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If global stereo matching methods are used to achieve high accuracy disparity maps, then measurement precision is improved, but computational complexity increases making hardware implementation difficult
Solution Approach 1:
The patent segments the image into superpixels based on dominant orientations before performing stereo matching. This segmentation groups pixels with similar orientation characteristics, reducing the number of independent matching operations while preserving important edge and texture information. The superpixel-based approach maintains measurement precision by ensuring that pixels within each superpixel share similar orientation properties, while significantly reducing computational complexity compared to processing each pixel individually.
Solution Approach 2:
The patent applies local quality by using dominant orientation detection to identify and weight pixels based on their local structural characteristics. Pixels located on dominant orientations (edges, contours) are given higher weights and processed with greater precision, while pixels in homogeneous regions are processed more coarsely. This selective approach maintains high accuracy for critical features while reducing overall computational burden, enabling practical hardware implementation.
2Productivity
If sparse-to-dense methods are used to reduce computational load, then productivity is improved, but error rates increase in texture-less regions or repetitive structures
Solution Approach 1:
The patent applies local quality by using dominant orientation detection to identify and weight pixels based on their local structural characteristics. Pixels located on dominant orientations (edges, contours) are given higher weights and processed with greater precision, while pixels in homogeneous regions are processed more coarsely. This selective approach maintains high accuracy for critical features while reducing overall computational burden, enabling practical hardware implementation.
Solution Approach 2:
The patent performs preliminary dominant orientation detection and superpixel segmentation before the actual stereo matching process. By pre-identifying regions with reliable structural information (dominant orientations) and grouping them into superpixels, the system prepares a refined input structure that guides the subsequent sparse-to-dense matching process. This preliminary action ensures that texture-less and repetitive regions are handled with appropriate weighting and grouping, reducing error rates while maintaining computational efficiency.
3Measurement precision
If all pixels are processed in stereo matching to ensure complete coverage, then measurement precision is improved, but use of energy increases making hardware implementation impractical
Solution Approach 1:
The patent segments the image into superpixels based on dominant orientations before performing stereo matching. This segmentation groups pixels with similar orientation characteristics, reducing the number of independent matching operations while preserving important edge and texture information. The superpixel-based approach maintains measurement precision by ensuring that pixels within each superpixel share similar orientation properties, while significantly reducing computational complexity compared to processing each pixel individually.
Solution Approach 2:
The patent applies partial action by selectively processing pixels based on their dominant orientation characteristics rather than uniformly processing all pixels. Pixels located on dominant orientations are processed with higher priority and precision, while pixels in homogeneous regions are processed more coarsely or grouped into larger superpixels. This selective processing approach ensures complete coverage of important features while reducing overall energy consumption to levels suitable for hardware implementation.
Data Source
AI summary
A method of matching stereo images includes: determining, first information indicating which orientations are dominant in each pixel location of a right image and a left image among the stereo images; determining second information indicating for each pixel location in the right image and the left image whether it lies on a pure horizontal edge; detecting points in the left image and the right image based on the first information and the second information; and performing a matching on the left image and the right image using the detected points to estimate a sparse disparity map.