Authomatic method based on image analysis for pilling evaluation in fabrics

Abstract

A fully automatic method for pilling evaluation in wear-andtear fabrics is developed from the image analysis of a set of standard photographs (Zweigle KG-741 Reutlingen). The method involves operations in both the spatial and frequency domains to segment pills from the textured background of the web. It calculates the total area of pilling in the sample image and assigns a degree of pilling according to the standard. Two mathematical descriptions are analyzed according to the underlying rule established by the standard images, using the visual estimation of the area of pilling performed by a group of observers. A logarithmic (in base two) approach, which is consistent with human visual perception laws and facilitates an optimization of the method, is eventually adopted.