Unsupervised texture segmentation based on redundant wavelet transform

Abstract

The algorithm of Redundant Wavelet Transform (RWT) and laws texture measurement is proposed and applied to image segmentation. Based on the characteristics of the indentation images, this article uses texture features to extract the indentation silhouette from the point view of texture segmentation. We adopt Redundant Wavelet Transform and laws texture measurement algorithm to describe the texture characteristics of the indentation image, forming a n-dimensional feature vector, introducing texture features smoothing algorithm based on quadrant to smooth the features. Finally we combine with the improved k-means clustering algorithm to get texture segmentation result. The experiment demonstrates that in the material Vickers hardness image segmentation the proposed algorithm was significantly effective and robust. © 2012 Springer-Verlag Berlin Heidelberg.