What is it about?
Considering the demand of reliable computation of matching cost in stereo image matching, a new weighted matching cost computation method based on nonsubsampled contourlet transform is proposed in this paper.
Featured Image
Photo by timJ on Unsplash
Why is it important?
One of the chief tasks in stereo image matching is computation of matching cost. The matching cost is the basis for finding conjugate pixels in two images taken at the same scene, and plays an important role in such fields as remote sensing image registration, 3D vision based on images, digital photogrammetry, etc. The robustness of the matching cost has a direct impact on reliability of matching results.
Perspectives
Considering the demand of reliable computation of matching cost in stereo image matching, a new weighted matching cost computation method based on nonsubsampled contourlet transform is proposed in this paper. It was tested and compared to other methods with internationally recognized Middlebury benchmark. Experimental results verify that the proposed algorithm has higher matching accuracy than traditional gray-based measures or feature-based measure such as NCC, SAD, SIFT, etc. Moreover, how the change of NSCT scale and direction parameters affects stereo matching accuracy is proved by the experimental analysis. Furthermore, the optimum values of NSCT parameters with optimal matching accuracy are confirmed in this paper.
Dr Ka Zhang
Nanjing Normal University
Read the Original
This page is a summary of: Stereo matching cost computation based on nonsubsampled contourlet transform, Journal of Visual Communication and Image Representation, January 2015, Elsevier,
DOI: 10.1016/j.jvcir.2014.10.002.
You can read the full text:
Resources
Contributors
The following have contributed to this page
