Defect detection in colored texture surfaces using filter..

This paper presents a Gabor filtering approach for automatic inspection of defects in colored texture surfaces. It simultaneously measures both chromatic and textural deviations of an image. Two brightness invariant chromatic features derived from different color spaces are used to form a complex number, which replaces the single gray-level information, for color pixel representation. The proposed method does not rely on the extraction of textural features. It is based on the energy response from the convolution of a Gabor filter with the color image characterized by the two-chromaticfeature complex number. This converts the difficult defect detection in a colored texture image into a simple binary thresholdng in the filter image. Experiments on a number of real colored texture surfaces such as textile fabric and wool have shown the effectiveness of the proposed method.