Abstract :
This paper presents a detailed study of different
clustering based image clustering algorithm. A cluster is
collection or group of data objects that are similar to each other
with the same cluster object and not similar with other cluster
object. Also it is study on different fuzzy rule based clustering
algorithm. To overcome the limitations of conventional FCM its
need to study Kernel fuzzy c-means (KFCM) algorithm in detail.
Basic K-means algorithm is sensitive to noise and outliers so, and
changes of K-means called as Fuzzy c- means (FCM) are
developed .FCM allows data points to belong to number of
cluster where each data point has own degree of membership of
belonging to each cluster. The KFCM uses a unique function and
gives better performance than FCM in case of noise corrupted
images. So it is nothing but grouping of set of physical data
objects into the classes of similar or matching objects. The fuzzy
rule clustering is the crisp clustering at the boundaries among
the cluster are vague and ambiguous. Up to yet the cluster never
can be identified by the human directly but which was possible
for the machines or system to identify cluster easily as per the
requirements of dataset or system. The cluster which is fuzzy in
nature is quite difficult to understand. The most drawback of
fuzzy and crisp clustering algorithm is there nature of sensitivity
to number of potential cluster and their initial position. The
image clustering is not easy to understand for human up to yet.
This is concept behind of this fuzzy clustering to make it possible
to understand for human, And also to make the crisp and
boundaries easy for the image cluster. The accuracy of the
finding image cluster should be to maintain with respective rate.
This will be another attempt to make it possible by using
different types of algorithm.
Keyword :
Clustering, Fuzzy, Boundaries, Initial, Crisp, Fuzzy Clustering, FCM, KFCM, NMKFCM.