magnetic resonance image segmentaion using k-means algorithm

in the context of magnetic resonance MR image segmentaion k-means clustering algorithm is used for segmentaion and compression two different distances which are often used in minimum distance classification algorithm are studied on Iris database it is shown that mahalanobis distance has a major drawbackfor automatic segmentation the robustness of the algorithm to noise versus overlapping intensities is also studied. it is shown that for segmentaion of 2-D slices of MR images which can later be used in intra-slice compression of 3-d MRimages only four classes are subjectively necessary.

کلیدواژه ها:

MR image segmentaion ، image compression ، clustering ، k-means algorithm ، Iris database