Learn how easy it is to sync an existing GitHub or Google Code repo to a SourceForge project! See Demo
I am using CStereoMatcher::Match method to find the correlated match in two digital images. The method sometime works well for one set of images and fails for the other. I have properly calibrated the webcamera but still there is inconsistency. The correlated point does lie on the epipolar line but sometimes it shows wrong match on the epipolar line. I have tried everything. I have also gone through the IVT help document givne on dOxygen. It's mentioned that the method depends on image material. I am confused what does this mean? I am trying to work out on this from so many days. I would be thankful to you if you could help me in this regard.
In addition to the query above, I would also like to know the difference between the StereoMatcher: Match method and StereoVision:Process. I would be thankful to you if you could help me in this regard.
stereo matching is often problematic in practice, and a good understanding of what is happening is very helpful. I recommend to read literature on this topic, e.g. our book "Computer Vision - Principles and Practice" (see ivt.sourceforget.net for details) gives an introduction, including an example application in chapter 11.
Regarding your questions:
1. If the stereo matching using stereo correlation works or not, depends on the image information that is available along the epipolar line. If there is enough information within the correlation window to allow to compute a unique matching, it will succeed, otherwise not. Also make sure to adjust the parameters d1, d2 so that the disparity interval is large enough to contain the searched disparity, but also not unnecessarily big, so that the likelihood of computing wrong matches is not increased too much.
2. CStereoMatcher computes matches for single points, whereas CStereoVision computes dense disparity maps. CStereoMatcher::Match uses the ZNCC (Zero-Mean Normalized Cross Correlation), CStereoMatcher::MatchZSAD the ZSAD (Zero-Mean Sum of Absolute Differences), and CStereoVision::Process and CStereoVision::ProcessFast the SAD (Sum of Absolute Differences).