How can I optimize 3000 variables for spline control points in image warping?

[physical101]

Dear All

Firstly thank you for looking at my post. I am trying to nonlinearly warp an image with respect to another using a free form deformation. I am trying to code the following paper in matlab:

http://www.cs.jhu.edu/~cis/cista/746/papers/RueckertFreeFormBreastMRI.pdf

I understand most of the implementation but was left confused about how to optimise the deformation of the spline control points. In the paper grid points are moved in order to minimise a mutual information metric, I am using simple overlap minimisation but the procedure should be the same. Can you really optimise 3000 variables? Or have I missed the point entirely.

Hoping you can help

Physical101

 

[Valkarie]

Yes, you can optimize 3000 variables. Generally, techniques such as gradient descent or other optimization algorithms are used to minimize the cost function. You may want to consider using a genetic algorithm, as this could be more efficient than gradient descent in this case. Additionally, you may want to try reducing the number of variables by clustering the control points, which will reduce the number of free variables.