erogol's profile - activity

I am also working through MATLAB and change the code OpenCV when it is ready. Sure using predefiend shape information is very useful idea but again the problem is, for each stadium we need to measure the distance to Camera to estimate the necessary shape of the final field view. Moreover, field scales are not constant and changing place to place. Therefore, predefining expected field shape then introduces additional hand made problems. But very helpful thread thanks... I also read the document you linked. :)

I tried many of the ordinary stitching procedures based on feature matching between the overlapping regions of both images and estimating Homography through those matches. However I could not get very accurate results even I tried different estimation algorithms like RANSAC, Norm 8 Points or others.

Another problem here, I should get a stitched image just like it is taken from a single camera at the mid side of the field. Therefore if I project for example first image to the other via estimated Homography it scewed the first image so the final image as well. I guess for good quality I need to project both images relative to other so that they match each other at some point.

Any suggestion or comment is very welcome...

Actually for each run ,even the pair is same, I get different stitching result because of estimation of H matrix is not very robust despite of RANSAC. I decided to use the frames of the videos to enhance the number of mathing interest points so as to have more points for RANSAC. I believe that RANSAC will be more robust with more precise H estimation. I am asking that is this true or how can I use the nature of video for better stitching.?

I try to stitch two views of the scene by estimating homography from overlapping segment. However, because of the nature of approximation of homography matrix, system gives slightly different results for each run. Then since I have same time videos of both views, I think about to use t0matching frames in time to have more robust estimation of H matrix. The question is, how can I get those multiple frames useful for the robustness of the system ?

I have one image and I want to recover its distortion by known camera distortion parameters k1 and k2. What should I do to rectify image view by those params?

I have two images taken from left angle and right angle of the view and I want to project those image into a exact frontal view, that is all vanishing point at the inf.

There is epipolar realtionbetween these two views. I can estimate the projection to get one from another. However, I need to manipulate this homography matrix to get the frontal view instead of pair-wise projection. How can I solve this problem?

Given the similar soccer field image, First, I need to detect the midfield line and more the merrier, I would like to learn the image projection for making the midfield line orthogonal evading the perfective distorion as if I am directly looking to the mid-field.

How can I apply those with OpenCV library?

I have a hard stitching problem therfore I need to define my own feature extraction routine instead of native SURF or other.

The main problem is, my images are symmetric to each other. Therefore, matching features are also symmetric that skwed the expected homography estimation. Hence I need to define the region that suposedly used in feature extraction. I also believe Lowe's SIFT is better than the SURF so I need to define my feature extraction routine using SIFT on a given area of the images.

How can I inject my own feature extraction routine, for instance well-known SIFT implementation of DAVID LOWE, instead of SURF at the below code that I work on ?

vector<Mat> images;
images.push_back(imread("images/isvec_left.jpg"));
images.push_back(imread("images/iscev_right.jpg"));

Mat pano;
Stitcher stitcher = Stitcher::createDefault(true);

stitcher.setWarper(new PlaneWarper());
stitcher.setFeaturesFinder(new detail::SurfFeaturesFinder(1000,4,4,4,4));
stitcher.setRegistrationResol(0.1);
stitcher.setSeamEstimationResol(0.1);
stitcher.setCompositingResol(1);
stitcher.setPanoConfidenceThresh(1);
stitcher.setWaveCorrection(true);
stitcher.setWaveCorrectKind(detail::WAVE_CORRECT_HORIZ);
stitcher.setFeaturesMatcher(new detail::BestOf2NearestMatcher(false,0.3));
stitcher.setBundleAdjuster(new detail::BundleAdjusterRay());
Stitcher::Status status = Stitcher::ERR_NEED_MORE_IMGS;
try{
    status = stitcher.stitch(images, pano);
    cout << status << endl;
}
catch(cv::Exception e){
    cout << "Stitching is not successfull!!" << endl;
}
imwrite("RESULT.jpg", pano);