Fabric pilling image segmentation based on mean shift

Abstract

Fabric appearance is always considered to be one of the most important aspects of fabric quality. Testing for fabric appearance is the process of inspecting, measuring and evaluating characteristics and properties of a fabric surface. Fabric Pilling is a key step in fabric pilling objective evaluation,which is the important component of textile performance test digitization.Image analysis has been widely accepted as an objetive mothod for evaluating fabric appearance.This study presents the principles of new method of fabric pilling image segmenttation based on mean shift.The principle of mean shift was demonstrated, and the extend principle of mean shift was educed. The extended mean shift algorithm was used to try to solve the segmentation of fabric pilling image.In this issue, two main steps were introduced: the filting of image and the segmentation of image. The influences of three parameters to the segmentation effect were analysised. The laboratory result shows that the proposed algorithm can get excellent segmentation after chosen three better parameters. © 2011 Springer-Verlag Berlin Heidelberg.