Function for elegantly performing range thresholding with arbitrary colors and ranges with OpenCV

colorThreshold.cpp

// This code is modified from the working version to remove coupling with the original
// application, and hasn't been tested in its new form. So, it's enough to give you
// the basic idea, but may not entirely work ;)
//
// It assumes the src image is in HSV and the dst image is grayscale.
//
//    CvScalar red = cvScalar(179, 255, 255);
//    CvScalar lowOffset = cvScalar(-20, -150, -150); // find colors down to h = 159, s = 105, v = 105
//    CvScalar highOffset = cvScalar(20, 0, 0); // find colors up to h = 19, s = 255, v = 255 (h wraps/is mod 179)
//    colorThreshold(img, thresholded, red, lowOffset, highOffset);
void function colorThreshold
(IplImage & src, IplImage & dst, CvScalar & color, CvScalar & low, CvScalar & high)
{
    cv::Mat m_src = src;
    cv::Mat m_dst = dst;

    int hMidPoint = 89;
    // you may want to change these or pass them in to threshold saturation or value
    int sMidPoint = 125;
    int vMidPoint = 125;
    int rotation = hMidPoint - color.val[0];

    for(int row = 0; row < m_src.rows; ++row) {
        uchar * p_src = m_src.ptr<uchar>(row);
        uchar * p_dst = m_dst.ptr<uchar>(row);
        for (int col = 0; col < m_src.cols; ++col) {
            int srcH = *p_src++;
            int srcS = *p_src++;
            int srcV = *p_src++;

            int dst = 0;

            int rotated = (srcH + rotation) % 179;
            if(rotated >= hMidPoint + low.val[0] && rotated <= hMidPoint + high.val[0] &&
               srcS >= sMidPoint + low.val[1] && srcS <= sMidPoint + high.val[1] &&
               srcV >= vMidPoint + low.val[2] && srcV <= vMidPoint + high.val[2]) {
                dst = 255;
            }
            *p_dst++ = dst;
        }
    }
}

Good contributor thanks,
For others can you init s an v

What should s and v be initialized to? And what do they represent? some type of saturation offset?

yes, if you were performing a saturation/value thresholding, they would be the midpoints of your range. I'll update the code.