OpenCV How to Plot velocity vectors as arrows in using single static image

Question

I am trying to plot velocity vectors like in matlab we use "quiver" function http://www.mathworks.com/help/techdoc/ref/quiver.html

I need to port same methodology in C++ using OpenCV library.

I have heard There are a few optical flow methods, i.e. Lucas and Kanade (cvCalOpticalFlowLK) or Horn and Schunck (cvCalOpticalFlowHS) or Block Matching method (cvCalOpticalFlowBM)

but all of these functions take two images , while i need to use one image because i am working on fingerprints.

Kindly help me ...

[Edit] Solution found

void cvQuiver(IplImage*Image,int x,int y,int u,int v,CvScalar Color,
                                            int Size,int Thickness){
cv::Point pt1,pt2;
double Theta;
double PI = 3.1416;

if(u==0)
    Theta=PI/2;
else
    Theta=atan2(double(v),(double)(u));

pt1.x=x;
pt1.y=y;

pt2.x=x+u;
pt2.y=y+v;

cv::line(Image,pt1,pt2,Color,Thickness,8);  //Draw Line


Size=(int)(Size*0.707);


if(Theta==PI/2 && pt1.y > pt2.y)
    {
    pt1.x=(int)(Size*cos(Theta)-Size*sin(Theta)+pt2.x);
    pt1.y=(int)(Size*sin(Theta)+Size*cos(Theta)+pt2.y);
    cv::line(Image,pt1,pt2,Color,Thickness,8);  //Draw Line

    pt1.x=(int)(Size*cos(Theta)+Size*sin(Theta)+pt2.x);
    pt1.y=(int)(Size*sin(Theta)-Size*cos(Theta)+pt2.y);
    cv::line(Image,pt1,pt2,Color,Thickness,8);  //Draw Line
  }
else{
    pt1.x=(int)(-Size*cos(Theta)-Size*sin(Theta)+pt2.x);
    pt1.y=(int)(-Size*sin(Theta)+Size*cos(Theta)+pt2.y);
    cv::line(Image,pt1,pt2,Color,Thickness,8);  //Draw Line

    pt1.x=(int)(-Size*cos(Theta)+Size*sin(Theta)+pt2.x);
    pt1.y=(int)(-Size*sin(Theta)-Size*cos(Theta)+pt2.y);
    cv::line(Image,pt1,pt2,Color,Thickness,8);  //Draw Line
    }

}

Answer 1

I am kind of completing the current answer here, which fails in giving the right size of each of the arrows' tip. MATLAB does it in a way that when an arrow is nearly a dot, it doesn't have any tip, while for long arrows it shows a big tip, as the following image shows.

To get this effect, we need to normalise the "tip size" of each of the arrow over the range of arrows' length. The following code does the trick

    double l_max = -10;

    for (int y = 0; y < img_sz.height; y+=10)                                                           // First iteration, to compute the maximum l (longest flow)
    {
        for (int x = 0; x < img_sz.width; x+=10)
        {
            double dx = cvGetReal2D(velx, y, x);                                                        // Gets X component of the flow
            double dy = cvGetReal2D(vely, y, x);                                                        // Gets Y component of the flow

            CvPoint p = cvPoint(x, y);

            double l = sqrt(dx*dx + dy*dy);                                                             // This function sets a basic threshold for drawing on the image

            if(l>l_max) l_max = l;
        }
    }


    for (int y = 0; y < img_sz.height; y+=10)
{
    for (int x = 0; x < img_sz.width; x+=10)
    {
        double dx = cvGetReal2D(velx, y, x);                                                        // Gets X component of the flow
        double dy = cvGetReal2D(vely, y, x);                                                        // Gets Y component of the flow

        CvPoint p = cvPoint(x, y);

        double l = sqrt(dx*dx + dy*dy);                                                             // This function sets a basic threshold for drawing on the image
        if (l > 0)
        {
            double spinSize = 5.0 * l/l_max;                                                        // Factor to normalise the size of the spin depeding on the length of the arrow

            CvPoint p2 = cvPoint(p.x + (int)(dx), p.y + (int)(dy));
            cvLine(resultDenseOpticalFlow, p, p2, CV_RGB(0,255,0), 1, CV_AA);

            double angle;                                                                           // Draws the spin of the arrow
            angle = atan2( (double) p.y - p2.y, (double) p.x - p2.x );

            p.x = (int) (p2.x + spinSize * cos(angle + 3.1416 / 4));
            p.y = (int) (p2.y + spinSize * sin(angle + 3.1416 / 4));
            cvLine( resultDenseOpticalFlow, p, p2, CV_RGB(0,255,0), 1, CV_AA, 0 );

            p.x = (int) (p2.x + spinSize * cos(angle - 3.1416 / 4));
            p.y = (int) (p2.y + spinSize * sin(angle - 3.1416 / 4));
            cvLine( resultDenseOpticalFlow, p, p2, CV_RGB(0,255,0), 1, CV_AA, 0 );

        }
   }
}

And this is an example of how this OpenCV code would look like

I hope this help other people Googling for the same issue.

Answer 2

Based on the code from Dan and the suggestion of mkuse, here is a function with the same syntax as cv::line():

static void arrowedLine(InputOutputArray img, Point pt1, Point pt2, const Scalar& color,
int thickness=1, int line_type=8, int shift=0, double tipLength=0.1)
{
    const double tipSize = norm(pt1-pt2)*tipLength; // Factor to normalize the size of the tip depending on the length of the arrow
    line(img, pt1, pt2, color, thickness, line_type, shift);
    const double angle = atan2( (double) pt1.y - pt2.y, (double) pt1.x - pt2.x );
    Point p(cvRound(pt2.x + tipSize * cos(angle + CV_PI / 4)),
    cvRound(pt2.y + tipSize * sin(angle + CV_PI / 4)));
    line(img, p, pt2, color, thickness, line_type, shift);
    p.x = cvRound(pt2.x + tipSize * cos(angle - CV_PI / 4));
    p.y = cvRound(pt2.y + tipSize * sin(angle - CV_PI / 4));
    line(img, p, pt2, color, thickness, line_type, shift);
}

We will see if those maintaining the OpenCV repository will like it :-)

Answer 3

The cvCalOpticalFlowLK does not plot velocity vectors, it computes these velocity vectors. If you do not have these vectors, you must call this function with two images. I guess you already have these vectors, and you just want to plot them.

In this case, you can use the cv::line function, for example:

cv::line(yourImage, cv::Point(baseX, baseY), cv::Point(endX, endY));

I hope this will help you!