Image Defect Detection Methods for Visual Inspection Systems

Two texture defect detection methods for automatic visual inspection systems will be presented in this paper. They divide up an analysed texture image into non-overlapping samples, and then calculate features of each sample using statistical analysis. Finally, the clustering of those features is applied to recognize the sample as defective or non-defective. Unlike the well-known methods, the proposed schemes do not require a previous training step to collect defective and non- defective texture samples. The experimental results show that these methods are effective and more accurate than earlier methods for image texture defect detection.