ANALYSIS OF GRAPH CLUSTERING BASED NORMALIZED GRAPH CUT FOR IMAGE SEGMENTATION

The humans have sense organs to sense the outside world. In these organs eyes are vital. The human eyes capture the light from the outside world and save the information as images in the brain. The human brain analyses the image data and gets the required information from the surroundings. Images are most prominent and easy way of representing a data. The art of representing information through the images is as old as the civilized man. Moreover the images can convey a clear data representation than the words or some other representation. Image segmentation is an old research topic, which has gained its importance in the past four decades. There are several previous methods for the segmentation. But there is no optimal solution for the judgment. This is because there is no specific benchmark for the judgment. In our project we propose a new method for the segmentation of an image called "The Normalized Graph Cut Segmentation". It is a global view concept which considers image as a graph model. The segmentation is done by using the similarity measurement technique. The problems of over segmentation and effect of noise can be overcome by this technique. The method is tested for various test cases like the landscape images, texture based images, high density feature based images and the performance of the algorithm has been tabulated.