Evaluation of image segmentation by kohonen neural network and k-mean cluster algorithm

Abstract EN

The present paper deals with image segmentation using Kohonen neural network and K-mean cluster approaches.

Image segmentation of the first technique has been implemented by using a single layer neural network trained by self-organization Kohonen competitive algorithm to produce a set of equiprobable weight vector.

These teehniques have been applied for three original images.

Kohonen approach gives better results with respect to K-mean cluster algorithm.

The reason for this enhanced result is that connectivity property between neighbouring in؛ the image is taken into consideration while its being neglected using k-means algorithm.

The connectivity property between neighbouring pixels is an important concept used in establishing boundaries of objects and components of regions in an image.