I'm new to the computer vision world. I'm trying to implement the Levenberg-Marquardt algorithm but I'm stuck at the computation of the jacobian. I'm trying to compute the jacobian of the 3d points from 2 views. I found this picture to help me implement it: jacobians What I do is for every point I concatenate this 2x4 matrix to form a "row". Then I append this rows toghether for every view. Now, I have a total of 2008 points and I get a matrix of 4x8032. When I try to compute the delta to update my 3d reconstruction I get a matrix of 8032 x 4016. Do you have any suggestions on what I am doing wrong?
I tried implementing the Jacobian in a different way, but at that point I can't get the multiplication between the Jacobian and the residuals.
