Rough-Fuzzy Clustering Algorithm for Segmentation of Brain MR Images - Rough Fuzzy Image Analysis: Foundations and Methodologies - page 32

means are shown in Figs. 2.5-2.7. All the results reported in Table 2.8 and Figs. 2.5-2.7

TABLE 2.8 Performance of Different C-Means Algorithms
DataSet Algorithms DBIndex DunnIndex β Index Time(ms)
I-204 HCM 0.15 2.64 12.44 4080
97761 FCM 0.12 2.69 13.35 38625
RCM 0.14 2.79 12.13 6670
RFCM 0.11 2.98 13.57 16532
I-204 HCM 0.16 2.03 13.18 3262
97763 FCM 0.15 2.24 13.79 45966
RCM 0.14 2.39 13.80 6770
RFCM 0.10 2.38 14.27 15457
I-204 HCM 0.16 2.38 8.94 3825
97777 FCM 0.15 2.54 10.02 40827
RCM 0.13 2.79 9.89 7512
RFCM 0.11 2.83 11.04 16930

confirm that although each c-means algorithm, except PCM and FPCM, generates good
segmented images, the values of DB, Dunn, and β index of the RFCM are better compared
to other c-means algorithms. Both PCM and FPCM fail to produce multiple segments of
the brain MR images as they generate coincident clusters even when they are initialized
with the final prototypes of other c-means algorithms.
Table 2.8 also provides execution time (in milli sec.) of different c-means. The execution
time required for the RFCM is significantly lesser compared to FCM. For the HCM and
RCM, although the execution time is less, the performance is considerably poorer than that
of RFCM. Following conclusions can be drawn from the results reported in this chapter:

FIGURE 2.5 I-20497761: segmented versions of HCM, FCM, RCM, and RFCM

FIGURE 2.6 I-20497763: segmented versions of HCM, FCM, RCM, and RFCM

1. It is observed that RFCM is superior to other c-means algorithms. However,
RFCM requires higher time compared to HCM/RCM and lesser time compared