Incorporating K-Means Clustering, DWT and Neural Network for Image Segmentation

Neelesh Singh Rajput, Shalini Sahay

Abstract: In the field of image segmentation, hybrid image segmentation techniques have always been a favorite way of researchers in past decades. In this paper we are going to propose a unique hybrid approach to image segmentation problem. Various images has been taken into this experiment to evaluate the proposed method. Features are extracted from the given image by using Discrete Wavelet Transformation (DWT) and Image gradient. Then the K-Means Clustering algorithm is fed with the features extracted which are unsupervised clustering method. Then the K-Means membership function is fed to the back propagating neural network as target value. Taking the features as input, Back propagation Neural network (BPNN) trained. Thus, to achieve a better solution to image segmentation problem, combination of K-Means Clustering and BPNN has been proposed in this paper. We have taken free images available in UCI Machine Learning repository.

Keywords: Image Segmentation, K-Means Clustering, Back Propagation Neural Network BPNN, Discrete Wavelet Transformation DWT