Texture Defect Detection System with Image Deflection Compensation

Image textural analysis technology has been widely used in the design of automated defect detection systems. Because the presence of defects may change the textural features of an image, a reference image without defects can be compared with the test image to detect whether there are any defects. However, besides defects, the deflection of the input test image could also change its textural features. When there is any angular difference between the reference and test images, their textural features would also be different, even if there is no defect in the test image. As a result, misjudgment of the defect detection system may occur. Most of the previous studies have focused on the development of textural analysis technology which could decrease the effect of test image deflection. This study aimed to estimate the deflection angle of test images through polar Fourier transform and phase correlation analysis, and rotate the reference image by the same angle to compensate for the deflection of the test image. After the angles of the reference and test images were brought into line, the textural analysis based on the gray level co-occurrence matrix was applied to analyze and compare the textural features of the two images. The results of actual texture defect detection demonstrated that the angular differences between the reference and test images could be estimated correctly, with an estimation error of only 0° to 0.5°. By compensating for the deflection of the test image, the accuracy of the texture defect detection could be effectively enhanced. Key-Words: Texture defect detection, Image deflection compensation, Polar Fourier transform, Phase correlation analysis, Gray level co-occurrence matrix