Segmentation_fuzzy_514_V2


Aim:
Perform gray level image segmentation on the basis of the gray level histogram, but without using the "traditional" gray level thresholding approach.

Method:
The method lies within the context of fuzzy methods.
First, the gray level histogram is computed.
Then, the histogram is smoothed in such a way that M modes are produced (M is the number of classes the segmentation will produce. It is fixed by the user.
On the basis of these M modes, grades of membership to the M classes are computed for each gray level.
M images, representing the grades of membership of every pixel to the M classes, are computed.
Probabilistic relaxation is applied to these images.
Finally, a defuzzification of the relaxed grades of membership produces the final segmented image.

The rationale for this method has been published in:

BONNET N., CUTRONA J. and HERBIN M.
    A �no-threshold� histogram-based image segmentation technique.
    Pattern Recognition (2002) 35(10), 2319-2322.
DOWNLOAD

V2 works on 3D images as well as 2D images.
Also: intra-class or (intra+inter-class) relaxation can be performed.


User interface (V2):

 
Number of classes: chosen by the user
Choice of a relaxation method: intra-class or (intra+inter-class) relaxation
Number of relaxation iterations: chosen by the user
Show the memberships; show the distances to modes; show the histograms::
    clicked ==> yes; unclicked (default value) ==> no


Illustration:

muscle
Original image

membership1_1 membership2_1 membership3_1
Grades of membership to the 3 classes (after 1 relaxation iteration)

membership1 membership2 membership3
Grades of membership to the 3 classes (after 5 relaxation iterations).

result_segm_fuzzy1 result_seg_fuzzy5
Final result (segmentation) after 1 and 5 relaxation iterations.