Possible bug with calcOpticalFlowPyrLK in version 2.4.1 (Bug #2075)

Could you attach a reproducing test or sample please?

I was able to find the source of the problem, the error occurs in my particular case, when the second parameter is a submat.
So something about the matrix being a submat or the fact that it is a non-continuous matrix produces that error.
I haven't tested with placing a submat on the first parameter or on both first and second, but be aware of it.

This crashes:

MatOfByte  status = new MatOfByte();
MatOfFloat err    = new MatOfFloat();
MatOfPoint2f currentPoints = new MatOfPoint2f();

// previousPoints is a non-empty MatOfPoint2f with some points    
// someMat and someSubmat have same type/size.

Video.calcOpticalFlowPyrLK(someMat, someSubmat, previousPoints, currentPoints, status, err);

With this work-around, it does not crash:

Mat testMat = someSubmat.clone();

// now testMat is not a submat and is continous

Video.calcOpticalFlowPyrLK(someMat, testMat, previousPoints, currentPoints, status, err);

Sure, I think this is what you are asking:

someSubmat.locateROI(wholeSize, ofs);

Log.d(TAG, "debug: "+someMat.size()+" "+someSubmat.size()+" / "+wholeSize.toString()+" / "+ofs.toString());

// Outputs:
// debug: 640x360 640x360 / 720x480 / {40.0, 60.0}