Math in computer vision

The importance of math in CV , CG or CP, is obvious. When we face a problem, math can be used to model and then explain the problem. When implementing these method practically, optimization is necessary, which can save the computation memory and speed up the computation process. Mathematical methods such as multiscale method(wavelet, Fourier transform etc.) , stochastic method (Bayesian method, MRFs, HMM, EM, Informax, Adboost, SVM, PCA, etc.), variational method( PDE, Level set, active contour, snake etc.) are explored by many researchers. Besides, many optimization methods, like Graph cut, dynamic programming, convex and non-convex optimization, are also used. In another point of view, these stuffs can be viewed as TOOLS when we solve a problem, explain one phenomenon or mechanism.
Some useful links are:
1. A seminar talking math methods in CV:http://www.cs.ualberta.ca/~vis/vision06/sessions.html
2. A seminar talking optimization methods in CV:http://visiontrain.inrialpes.fr/?page=school1
3. A course viewing CV in a mathematical way: http://www.cse.psu.edu/~cg586/