Detection of Localized Defects in Regular Textures with Simple and Complex Patterns
In the present paper, an algorithm is proposed for detection of localized defects in simple and complex texture patterns that are often found in regular textures like textile garments and wallpapers. The proposed method involves extraction of texel, reconstruction of an ideal, defect-free texture, linear filtering of error image obtained from original image and reconstructed image, and Canny edge detection on the error image. An ideal defect-free texture is created by tiling of mean texel extracted from the original texture using thresholding of autocorrelation function of the original image. Using the error image obtained from the difference between the original texture and the reconstructed texture, localized defects were identified and located with the help of linear filter and Canny edge operator.