Title: Robust and automatic segmentation of a class of fuzzy edge images

Authors: ZhenZhou Wang, YuMing Zhang

Addresses: Department of Electrical and Computer Engineering, University of Kentucky, Lexington, Kentucky 40506, USA. ' Department of Electrical and Computer Engineering, University of Kentucky, Lexington, Kentucky 40506, USA

Abstract: This paper concerns the segmentation of a class of images which consist of a background region, a fuzzy/blurry edge region, and an object region. In this class of images, the greyness gradually changes from the background region to the object region. To appropriately segment this class of images, the authors propose to model the fuzzy edge region using double-thresholds. In addition, a probability is assigned to each pixel in the fuzzy edge region for its membership to the object based on the differences between its greyness to the thresholds. To be robust, a statistical method, namely the maximum slope difference principle, is used to obtain optimal estimates of the thresholds automatically. Metal transfer images acquired from gas metal arc welding are used to demonstrate the algorithm and its effectiveness.

Keywords: image processing; gas metal arc welding; manufacturing; robust segmentation; automatic segmentation; fuzzy edge images; maximum slope difference; GMAW.

DOI: 10.1504/IJMIC.2011.037835

International Journal of Modelling, Identification and Control, 2011 Vol.12 No.1/2, pp.88 - 95

Published online: 31 Dec 2010 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article