Color Image Segmentation Using Improved Region Growing and K-Means Method

Abstract

-In areas such as computer vision and image processing, image segmentation has been and still is a relevant research area due to its wide spread usage and application. The traditional segmentation technique which is used in gray-scale mathematical morphology is watershed transform. Region Growing is an approach to image segmentation in which neighbouring pixels are examined and added to a region class if no edges are detected. This process is iterated for each boundary pixel in the region. In this paper, we made enhancements in watershed algorithm and region growing algorithm for image and color segmentation. The new enhanced algorithm is implemented in MATLAB and results are compared with the existing technique in the form of visualization and on the basis of Liu's F-factor values. I. INTRODUCTION Image segmentation is the process of dividing the given image into regions homogenous with respect to certain features, and which hopefully correspond to real objects in the actual scene. Segmentation plays a vital role to extract information from an image to create homogenous regions by classifying pixels into groups thus forming regions of similarity. The homogenous regions formed as a result of segmentation indwell pixels having similarity in each region according to particular selection criteria e.g. Intensity, color etc. Image segmentation is the basic requirement of any computer vision application because people are generally interested only in certain parts of the image. Image segmentation results in non overlapping objects labeled with different region numbers. It should be noticed that no general technique has been developed yet to segment an image precisely, so different techniques are taking floor to perform segmentation. Threshold based image segmentation techniques discriminate regions on the basis of intensity value difference between pixels. Thresholds for image segmentation have been calculated based on maximum entropy, interclass variation, and histogram. The limitation of threshold based segmentation technique is that it performs well for images, which have only two components. For complex images, it is calculated to support further processes. Clustering is an approach in which pixels are classified to a cluster, which is closest among all clusters. Pixels having homogeneous characteristics belong to the same cluster and different with respect to pixels of other clusters. The pixels must follow the homogeneity criteria in the same cluster. In cluster based image segmentation techniques, it is necessary to choose a certain number of clusters initially which eventually reduces the dynamicity of the technique.