Conceptually it is similar to the classifiers except that they are implemented in the spatial domain of an image rather than in a feature space. It has been widely used for MRI brain image segmentation of various structure as well as the brain volume from head scan. It treats the segmentation as a registration process. Some researchers used atlases not only to impose spatial constraint but also provide probabilistic information about the tissue model. They proposed a probabilistic tissue model and used expectation-maximization [23] to segment brain tumor by modifying an atlas with patient specific information about tumor location from different MRI a modalities [24][25]. The advantage is that it has ability to segment an image with no well-defined relation between regions and pixels.
K-means is a clustering method that partitions the n-points into the k-clusters in which each pixel belongs to one cluster by minimize an objective function in such a way that with in a cluster sum of squares is get minimized. It starts with k-clusters and each pixel is assigned to one cluster. T. U. Paul and S. K. Bandhyopadhyay [26] have proposed a two-way step segmentation process of brain MRI image. The limitation of K-means algorithm is computational time increases on implementing in large amount of data. On using a large amount of data set the storage space increases in K-mean …show more content…
Fuzzy clustering a soft segmentation method has been widely used in image clustering and image segmentation. Among the clustering method FCM [28] is the most widely used for image segmentation because it can retain more information than the hard segmentation methods [29].This method works well for assigning the membership each data point corresponding to each cluster center on the basis of difference between cluster and data point. It is better than k-means where data point belongs to one cluster center but in FCM data point belong to more than one clusters. The application of FCM to MR data has shown encouraging results .therefore, FCM for segmenting brain tumors is becoming a fruitful research area but it does not consider any information about spatial context. Thus, many researchers has incorporated local spatial information into the original FCM algorithm in order to improve the performance of image segmentation. Tolias and panas [30] developed a rule-based neighborhood enhancement system to impose spatial constraints by post processing the FCM clustering result. Pham [31] modified the FCM objective function by including a spatial penalty on the membership function. Ahmed et al proposed the FCM_S where the objective function is modified in order to compensate the intensity inhomogeneity. One of the disadvantage of FCM_S is a very time consuming process. In order to reduce the computation speed Chen