Abstract: The Gabor filter was applied to the strip surface defect detection because of its selectivity of frequency and direction. In addition to its denoisability, the detection system can keep the images of textural characteristics of defects completely. With the evaluation function introduced in, the difference between energy responses from both defective and non-defective images was maximized to determine the optimum parameter setting of the filter. Based on the optimum threshold function, the binarized images were obtained and the low-level noises involved in those images were removed by morphological analysis method, thus providing the segmented defective images. Experimental results verified the effectiveness of the method proposed.