Image Stitching Opencv

In an effort to produce an image stitching program, I have used this tutorial in order to write the following program:

void stitchImages(Mat imageOne, Mat imageTwo) { 
    namedWindow("Result", WINDOW_AUTOSIZE);
    Mat firstImageDescriptors, secondImageDescriptors;

    Mat imageOneGray, imageTwoGray;
    
    cvtColor(imageOne, imageOneGray, CV_RGB2GRAY);
    cvtColor(imageTwo, imageTwoGray, CV_RGB2GRAY);

    cv::Ptr<Feature2D> featureDetector = xfeatures2d::SURF::create();
    
    vector <KeyPoint> firstImageKeyPoints, secondImageKeyPoints;
    cout << "About to detect" << endl;
//    featureDetector->detect(imageOneGray, firstImageKeyPoints);
  //  featureDetector->detect(imageTwoGray, secondImageKeyPoints);

    DescriptorExtractor descriptorExtractor;
    cout << "About to extract" << endl;
 //   descriptorExtractor.compute(imageOneGray, firstImageKeyPoints, firstImageDescriptors);
   // descriptorExtractor.compute(imageTwoGray, secondImageKeyPoints, secondImageDescriptors);

    featureDetector->detectAndCompute(imageOne, Mat(), firstImageKeyPoints, firstImageDescriptors, false);
    featureDetector->detectAndCompute(imageTwo, Mat(), secondImageKeyPoints, secondImageDescriptors, false);

    cout << "About to Flann it up" << endl;

    FlannBasedMatcher matcher;
    vector <DMatch> matches;
    matcher.match(firstImageDescriptors, secondImageDescriptors, matches);


       double max_dist = 0; double min_dist = 100;

    for(int i = 0; i < imageOneGray.rows; i++) {

        double dist = matches[i].distance;
        
        if(dist < min_dist) min_dist = dist;
        if(dist > max_dist) max_dist = dist;

    }

    vector<DMatch> good_matches;

    for(int j = 0; j < imageOneGray.rows; j++) {

        if(matches[j].distance < 4*min_dist)
            good_matches.push_back(matches[j]);

    }
 Mat img_matches;
//drawMatches(imageOne, firstImageKeyPoints, imageTwo, secondImageKeyPoints, good_matches, img_matches);
 //   imshow("Result", img_matches);
  //  waitKey(0);


    vector <Point2f> imageOnePoints;
    vector <Point2f> imageTwoPoints;

    for(int k = 0; k < good_matches.size(); k++) {

        imageOnePoints.push_back(firstImageKeyPoints[good_matches[k].queryIdx].pt);
        imageTwoPoints.push_back(secondImageKeyPoints[good_matches[k].trainIdx].pt);

    }
   
    Mat homographyMatrix = findHomography(imageOnePoints, imageTwoPoints, 8);

    Mat finalImage;
    warpPerspective(imageOne, finalImage, homographyMatrix, Size(imageOne.cols + imageTwo.cols, imageOne.rows));
    imshow("Result", finalImage);
    waitKey(0);
    Mat half(finalImage, Rect(0, 0, imageTwo.cols, imageTwo.rows));

    imageTwo.copyTo(half);

    

    imshow("Result", finalImage);
     waitKey(0);

    //set work_megapix
    //set work_scale

    }

(Please ignore the commented out code and notes to self)

I am using OpenCV 3.0 while this one uses a version of 2.0 to the best of my knowledge, so I had to change some things. I originally tried to stitch to images together, but the program failed. I was finding a decent amount of matches (I checked), but for some reason the transformation using the homography matrix I feel is the most likely culprit. I thought to myself, "this tutorials seems to be for panoramas without camera rotation, and the images I took are meant for a rotated panorama, so I might as well try taking some images where the camera does not rotate". I tried to do so and the results ended up with one image being copied to one half of the final pano and the other half of the pano being a gray smear. At least when I messed around with the rotated images, the first half of the image looked like a rotated version of the first image, yet the part that would have made it a panorama and not a copy of the second image's left part was cut-off, so I am not sure if it was just part of the second image (I feel it was probably legitimately part of the first). I am sorry I cannot post images, the image files for the final panoramas are too big. Any help would be much appreciated though. Thank you,

                                             - Jacob

P.S. I had to insert my code as a javascript/css/html code snippet so it would format correctly, but it obviously won't run in the browser so don't try! Also, I know I have two imshows that are not commented out at the end. I have two to see the final image looks before and after the "second image" is copied to the panorama.

via Chebli Mohamed