A survey of K means Clustering with modified gradient magnitude region growing technique for lesion segmentation

Image segmentation is very useful in lesion segmentation which is needed for monitoring and quantifying Lesion. Automated lesion segmentation in CT/MRI poses many challenges with regard to characteristics of an image. Many techniques are involved to detect the lesion in the medical images such watershed, wavelet transform etc. In this paper, we are defining the modified k means cluster to segment the lesion and combining the technique with Modified gradient magnitude region growing technique. Integrated techniques are analyzed as better for lesion segmentation. Keywordssegmentation, K mean clustering, region growing. I. INRODUCTION In computer vision, Segmentation refers to the process of partitioning a digital image into multiple regions (sets of pixels). The goal of segmentation is to simplify and/or change the representation of an image into something that is more meaningful and easier to analyze. Image segmentation is typically used to locate objects and boundaries (lines, curves, etc.) in images. Partitioning of an image into several constituent components is called segmentation. Segmentation is an important part of practically any automated image recognition system, because it is at this moment that one extracts the interesting objects, for further processing such as description or recognition. Segmentation of an image is in practice the classification of each image pixel to one of the image parts. Segmentation subdivides an image into its constituent regions or objects. The level to which the segmentation is carried depends on the problem being solved. That is, segmentation should stop when the objects of interest in an application have been isolated. For example in the automated inspection of electronic assemblies the interest lies in analyzing images of products with objective of determining the presence or absence of specific anomalies, such as missing components or broken connection paths II. IMAGE SEGMENTATION TECHNIQUES The most important step in Image Segmentation is to generate a compact description of an Image. In order to do that following two approaches could be used a) Contour Segmentation (edge detection and contour tracing), b) Region Segmentation (Grouping of connected pixels in to regions of uniform properties)We propose, in this paper, Region Segmentation Algorithm based on a edge preserving smoothing filter, the symmetric nearest neighbor mean and a fast anisotropic diffusion. There are two techniques of Image Segmentation – 2.1. Edge Preserving Smoothing Filter [4] Edge preserving smoothing filters cog nominate edge preserving noise-cleaning filters. Systematic Nearest Neighbor (SNN) filters, in many of the correlative studies of edge-preserving noise smoothing techniques, is considered to give the best results in smoothing and preserving edges. Therefore, SNN is most widely used as an Image Improvement Technique. Westman et al has used a SNN filter to preprocess color images before segmenting them. The principle on which the SNN filter works is stated below The SNN filter makes use of both the Spatial and Gray value information in the neighborhood of pixel to be processed. In a squre window, half the number of Pixels is selected by choosing one pixel nearest in Gray value to the center pixel from each pair of pixels located symmetrically opposite the center. Only the selected pixels are used to compute a new value for the center pixel. 2.2. Anisotropic diffusion [4] In Image processing, anisotropic diffusion is also called Perona–Malik diffusion. It is used to remove the Image Noise without losing the significant parts of the image content and other details that are important for the Image interpretation. This technique enhances the contrast by using a modified heat diffusion equation. This technique is a discontinuity preserving smoothing approach and is closely related to the adaptive smoothing proposed by Chen et al. The principle is that a pixel should become weighted average of its neighbors. The weight resembles the continuity measures of these pixels. Repetitive Implementation of anisotropic diffusion is Adaptive smoothing in which the unwanted edges will disappear along with repetition. However, this scheme is considered to be slow, and to avoid this slowness we use Toboggan Contrast enhancement as proposed by Fairfield. 2.3. Application of Edge Preserving Smoothing and Anisotropic diffusion [4] SNN filters are considered good for Clearing the noises and preserving the edge, but they cannot make potential regions Uniform. Whereas, anisotropic diffusion can make potential regions uniform but it is impressionable to noises. So, we propose to make a collective use of both the techniques to enhance an image before segmenting it. To make Fairfield’s diffusion algorithm less impressionable to noises, Canny-Deriche detector can be used by us. III. CLUSTERING Clustering is a collection of objects which are “similar” between them and are “dissimilar” to the objects belonging to other clusters. Therefore, it deals with finding a structure in a collection of unlabeled data. The goal of clustering is to determine the intrinsic grouping in a set of unlabeled data. Clustering Techniques can be divided into two categories3.1. Hierarchical Clustering [12] Hierarchical clustering techniques are based on the use of a proximity matrix indicating the similarity between every pair of data points to be clustered. The end result is a tree of clusters called a dendrogram representing the nested grouping of patterns and similarity levels at which groupings change [A.K. Jain and P.J.Flynn, 1999]. It proceeds successively by either merging smaller clusters into larger ones (agglomerative, bottom-up), or by splitting larger clusters (divisive, top-down). The clustering methods differ in regards to the rules by which two small clusters are merged or a large cluster is split. Some of the Hierarchical [R.Bhowmik, 2003] algorithms include COBWEB, CURE and CHAMEL. 3.2. Partitional Clustering Partitional algorithms are categorized into Partitioning Relocation Algorithms and Density-based Partitioning. Partitional clustering techniques such as K-means clustering have an advantage over the hierarchical clustering techniques, where a partition of the data points which optimizes some criterion functions. In hierarchical clustering once a “thing” is assigned to a particular cluster, it cannot be altered. Therefore if a “thing” is incorrectly assigned to a particular cluster at an early stage there is no way to correct the error. There is however a disadvantage of the partitional clustering techniques and this is the determination of the number of clusters K. IV. K-MEANS CLUSTERING ALGORITHM There are two existing basic versions of k-means clustering, a non-adaptive version introduced by Lloyd and an adaptive version [14] introduced by MacQueen. The most commonly used k-means clustering is the adaptive kmeans clustering based on the Euclidean distance. Adaptive k-means clustering can be considered as a special case of the gradient descent algorithm where only the winning cluster is adjusted at each learning step. This paper concentrates only on adaptive k-means clustering as the algorithm can be used for on-line training of RBF network. Adaptive k-means clustering tries to minimize the cost function in equation (1) by searching for the centre cj on-line as the data are presented. As the data sample is presented, the Euclidean distances between the data sample and all the centers are calculated and the nearest centre is updated according to: