Textile Inspection Based on À trous Wavelet Transform

In this paper, a textile defect detection method is introduced based on à trous wavelet transform. The processing procedure has two stages: training and testing. In each stage, the images are decomposed using à trous wavelet transform, and the approximate sub-bands are extracted at an appropriate level. This transform is a time-invariant transform. It is useful for several applications, such as textile defect detection. The purpose of the proposed method is to reinforce the energy of the defected area and reduce the energy of the background to allow the detection process, efficiently. The threshold values are energy estimations for rows and columns. The method is applied on different datasets; such as dataset 4 of DAGM dataset, the dataset from industrial automation research laboratory and our own practical dataset.