|
(1) Input: MR image |
(2) Output: Segmented tumor image |
(3) Preprocessing: Perform adaptive Wiener filter and |
morphological operation. |
(4) Set the value of clusters k, the degree of fuzziness , the |
error , and the value of objective function |
(5) Initialize the cluster centroid using K-means++: |
(6) Choose an initial center at random from image R, where x represents the pixel of the MR image |
(7) Begin |
(8) Calculate the probability of each remaining pixel using |
, where is the distance between the |
pixel and the nearest center, ā represents the next pixel. |
(9) Choose the pixel with the highest probability as the next |
initial center |
(10) Update the the i-th cluster centroid |
(11) If k initial centers are calculated |
(12) Then Break |
(13) End if |
(14) Cluster the obtained images using K++GKFCM: |
(15) Begin: |
(16) Calculate K(x, y) using Eq. (9) |
(17) Update the membership degree using Eq. (5) |
(18) Update the the i-th cluster centroid using Eq. (4) |
(19) If , where represents |
the function of the i-th iteration |
(20) Then Break |
(21) End if |
(22) End |
|