File Exchange

image thumbnail

stereo matching using adaptive random walk

version 1.2.0.0 (1.03 MB) by Sehyung Lee
The source code for stereo matching using random walk algorithm

4 Downloads

Updated 13 Jul 2017

View Version History

View License

"Robust Stereo Matching using Adaptive Random Walk with Restart Algorithm,"
Sehyung Lee, Jin Han Lee, Jongwoo Lim, Il Hong Suh, Image and Vision Computing. ( Accepted, Jan 22, 2015)
Recent online publication can be found at
http://www.sciencedirect.com/science/article/pii/S0262885615000104
----------------------------------------------------------------------------
Our algorithm is implemented in MATLAB with MEX.
Procedures
> run "compile_mexfiles.m", to complie MEX sources.
> run "run_matching.m"
Youtube video demonstration
https://www.youtube.com/watch?v=Cn7tLsWD8ws


Enjoy it!
----------------------------------------------------------------------------
Any questions or bugs to Sehyung Lee
shl@incorl.hanyang.ac.kr
2.Feb.2015

Cite As

Sehyung Lee (2020). stereo matching using adaptive random walk (https://www.mathworks.com/matlabcentral/fileexchange/49501-stereo-matching-using-adaptive-random-walk), MATLAB Central File Exchange. Retrieved .

Comments and Ratings (5)

Sehyung

Hi Ian, I'm sehyung, sorry for very late response because I moved to other university after graduation.
When I did this work, it was based on KITTI 2012 stereo dataset that was maybe different to yours. Also, in my paper, disparity maps such as Fig.2 and Fig.8 estimated on KITTI images were dense results estimated by random walk in order to check all pixel's accuracy. If you check the accuracy with like this way, you can get the same results with ours. Also we tested our method on KITTI benchmark with Out-Noc in which KITTI server interpolated disparities on these occlusion areas with their default method. So, denseity was recorded 99.33%.

Ian

hi,Sehyung Lee.I have a question,when I run "run_matching.m"(as your parameters in paper),now I get 194 left_disparity_map about tain set.then compare with the ground truth ,I get the noc err is 6.5%,occ err is 8.3%.I don't know why , because your result should be interpolated then compare with the ground truth or other?Besides,I watch from your paper,first ,I think the disparity is dense by your method?why the disparity is sparse?99.33% in test.second,why don't use the max disparty is 256 as the kitti dekit recommend.please answer me ,thank you very much .

Sehyung Lee

Thanks for the comment. I'll add an explanation about the role of each parameter as soon as possible in "run_matching" script.

Igor Varfolomeev

Indeed more robust than other methods in many cases. But the "run_mataching" script is poorly commented - it would be nice if there would be a short description of what each parameter means, and the units used.

Gaochang Wu

MATLAB Release Compatibility
Created with R2013b
Compatible with any release
Platform Compatibility
Windows macOS Linux

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!