I need to implement Expectation Maximization through the use of Matlab for image segmentation. The basic idea is below from previous work done. A framework of codes and a possible set of EM codes can be provided but modifications need to be done.
Not only would I need a working set of codes but I'll need guidance to understand both EM and the codes fully.
A sequence of images(eg. 20 frames) input into the framework codes which make use of EM to produce background models.(eg. 3 background models each with a mean and variance)
A current(test frame) image is being compared with the background models.
3 D-matrices are produced using D=I-mu/sigma (something like this) and the min is taken :
D=min(D(:,:,1), D(:,:,2), D(:,:,3));
Thresholding is done next (i.e. map=<5)
Blobs are formed for connected component analysis.
Small regions are removed for better performance.
Conversant with the EM algorithm and Gibbs sampling to compute maximum likelihood estimates given incomplete samples. 1 year experience in working with MATLAB. Research publication in 3D Image processing.