Defect Detection in Textile Fabrics using Modified Error Prediction Filters

In this paper, a filtering approach for defect detection in textile fabrics is presented. In this method, brightness of each pixel is approximated by linear combination of its surrounding pixels brightness. By using least square errors method, coefficients of the filter are approximated and then an error image will be constructed. By applying different types of filter structure and using a bi-level Threshold technique, it is shown that a wide range of irregularities in fabric texture can be detected and localized.

Keywords:

Fabrics , defect detection , error prediction , error filter , texture , bi-level threshold , machine vision