What matrix to use to estimate homography matrix in planar homography

If I have 4 point correspondences, eg: (x1,y1), (x2,y2), (x3,y3), (x4,y4), and I want to compute the H matrix.

I create the A matrix using the 4 points and have a 8*9 matrix.

I calculate SVD using [U, D, V] = svd(A).

Do I use U or V as the eigen vectors of the smallest eigen value to create H. How to create H matrix here?

Answer 1

[Matt J]

H=reshape(V(:,end),3,3);

[Matt J]

It would be better, however, to use fitgeotform2d.

[sarvan_keep]

Thank you. I understand. I am new to Matlab. Since matlab arranges matrices column-wise, do I not have to take a transpose to ensure the order of the H matrix? That is when I later use it to multiply it to some points of my choice.

[Matt J]

If you apply H using a right-side multiplication H*x, as opposed to a left-side multiplication x.'*H, then yes, you will have to transpose.