Optimized Gabor filter parameters for uniform texture flaw detection

Abstract

The problem of automated defect detection in uniform textured materials is investigated. A new approach for selecting Gabor filter parameters is proposed. The optimized odd symmetric Gabor filter is designed to match with the texture features of defect-free fabric image. The performance of the scheme is evaluated by using an offline test database with 60 uniform texture images. The experimental results exhibit accurate flaw detection with low false alarm. And the effectiveness and robustness of the proposed method are confirmed.