Revision history [back]

So you have a bunch of images and want to arange them in one large image. Let's say you put all your images first to a vector (images) and you want a specific number of columns (cols) and maybe some gap between the images (min_gap_size) then you can arrange them with the following method:

cv::Mat createOne(std::vector<cv::Mat> & images, int cols, int min_gap_size)
{
    // let's first find out the maximum dimensions
    int max_width = 0;
    int max_height = 0;
    for ( int i = 0; i < images.size(); i++) {
        // check if type is correct 
        // you could actually remove that check and convert the image 
        // in question to a specific type
        if ( i > 0 && images[i].type() != images[i-1].type() ) {
            std::cerr << "WARNING:createOne failed, different types of images";
            return cv::Mat();
        }
        max_height = std::max(max_height, images[i].rows);
        max_width = std::max(max_width, images[i].cols);
    }
    // number of images in y direction
    int rows = std::ceil(images.size() / cols);

    // create our result-matrix
    cv::Mat result = cv::Mat::zeros(rows*max_height + (rows-1)*min_gap_size,
                                    cols*max_width + (cols-1)*min_gap_size, images[0].type());
    size_t i = 0;
    int current_height = 0;
    int current_width = 0;
    for ( int y = 0; y < rows; y++ ) {
        for ( int x = 0; x < cols; x++ ) {
            if ( i >= images.size() ) // shouldn't happen, but let's be safe
                return result;
            // get the ROI in our result-image
            cv::Mat to(result,
                       cv::Range(current_height, current_height + images[i].rows),
                       cv::Range(current_width, current_width + images[i].cols));
            // copy the current image to the ROI
            images[i++].copyTo(to);
            current_width += max_width + min_gap_size;
        }
        // next line - reset width and update height
        current_width = 0;
        current_height += max_height + min_gap_size;
    }
    return result;
}

I hope it is clear what you have to do (in your do-while-loop you add all your matrices to a vector and afterwards you call createOne()).

Also note that you kinda mix the C-API of OpenCV with the C++ one, I recommend to switch completely to C++, i.e. imread() instead of cvLoadImage(), replace cvCreateMat to the proper Mat's and what does actually vconcat() do?