Markov random fields and Karhunen-Loeve transforms for defect inspection of textile products

In this paper the problem of using machine vision in quality inspection of textile fabrics is considered. A model based approach with Markov random fields (MRF) as the texture model and a new method based on Karhunen-Loeve transforms is studied for the defect inspection of textile fabrics. The results are illustrated on real fabric images, and the real time implementation of the MRF approach on a two TMS320C40 based parallel processing system is given.