EM-MPM Image Segmentation Algorithm

Segments a grayscale image using the EM-MPM algorithm.

Is = emmpm(I) returns an uint8 matrix Is representing the segmented image. Each entry of Is is an integer between 1 and 'regions'. 'regions' defines in how many regions the image I will be segmented.

EMMPM defines optional arguments:

EMMPM(I, regions, steps, mpmSteps, coolMax, coolInc)

steps: Defines the number of iterations in the segmentation process (default = 5);
mpmSteps: Defines the number of repetitions of the MPM algorithm at each iteration of the segmentation process (default = 1);

The EM/MPM method is iterative that stabilizes after a number of iterations. Usually, a higher number of iterations yelds better segmentation but increases segmentation time. Use a lower values for 'steps' and 'mpmSteps' to reduce time.

coolMax: Defines the maximum cooling rate (default = 1.2);
coolInc: Defines how much the cooling rate is incremented at each iteration (default = 0.025).

coolMax and coolInc control the segmentation cooling rate. Higher cooling rates stabilize the segmentation process. In early stages of segmentation, it is desired to have a low cooling rate.

Example:

% Binary segmentation.
I = imread('coins.png');
Is = emmpm(I, 2);
imshow(Is, []); % Show the segmented image.

% 5 regions segmentation.
I = imread('peppers.png');
Is = emmpm(I, 3);
imshow(Is, []); % Show the segmented image (note that emmpm does not use color information).


Example image from radiopaedia.org