Transformations with Emgu (OpenCV) - Affine/perspective?

Question

I'm currently trying to implement transformations through using EMGU, though I can't seem to get my head round how it works (and there doesn't seem to be any examples online).

I've got my image, with the 4 points I wish to transform from (and to), though I don't know what other variables are required, it asks for 'mapMat' ?

Here is what I have so far:

        float[,] tmp = {
                            {bottomLeft.x, bottomLeft.y},
                            {topLeft.x, topLeft.y},
                            {topRight.x, topRight.y},
                            {bottomRight.x, bottomRight.y}
                       };
        Matrix<float> sourceMat = new Matrix<float>(tmp);
        float[,] target = {
                            {0, height},
                            {0, 0},
                            {width, 0},
                            {width, height}
                       };
        Matrix<float> targetMat = new Matrix<float>(target);
        //mapMat = just a placeholder matrix?
        Matrix<float> mapMat = new Matrix<float>(target);
        CvInvoke.cvGetAffineTransform(sourceMat.Ptr, targetMat.Ptr, mapMat.Ptr);

This however doesn't work. I was also unsure whether or not an affine transformation was the most ideal solution? I read something about FindHomography too and also perspective transformations, but not sure if they would apply here.

The target transformation i wish to achieve is like this:

http://img832.imageshack.us/img832/5157/targettransform.png

Any help would be greatly appreciated,

Thanks

Answer 1

First a little intro:

• If you want an affine transformation (as it seems from your picture) you need at least 3 point pairs because you have 6 parameters to estimate
• If you want a more generic transformation, i.e. an homography you need at least 4 point pairs because you have 8 parameters to estimate

So suppose you have your 4 source and destination corners, and you want to estimate a Perspective Transformation this code should do what you want:

PointF[] pts1 = new PointF[4];
PointF[] pts2 = new PointF[4];
HomographyMatrix homography;
for (int i = 0; i < 4; i++)
{
  pts1[i] = new PointF (sourceCorner[i].X, sourceCorner[i].Y);
  pts2[i] = new PointF (destCorner[i].X, destCorner[i].Y);
}
homography = CameraCalibration.GetPerspectiveTransform(pts1, pts2);

Take a look at CameraCalibration for other useful methods