Abstarct: According to this fact that the inspection process for detecting defects in textile fabrics mainly need to be done on-line, process time for used algorithm is essential in its excitability. Most previous algorithms have not this capability because of heavy computing. Moreover, in real application, many defects in the fabric are unpredictable and have no clear structure. Thus training of system by having all type of defect practically is impossible. So defect detect system should be automated as much as possible and not require to training defective class. In this thesis a new approach is presented which remove training phases by using clustering instead of classification and detect the structure of textile fabric using only similarities and differences of data points in assumed databaxse. Fuzzy baxsed clustering has been our baxse of works and the genetic algorithm and wavelet transform is used. In the proposed method, we have been used wavelet transform as preprocessing to reduce the background of texture and clarify defects and FCM clustering algorithm to detect defects of patterned fabrics. Genetic algorithm covers a wider range of defects in patterned fabrics by acquire the optimal cluster centers and improving the performance of fuzzy clustering. Presented algorithm is able to implement as automatic defect detection system for texture of fabric by reducing computational complexity.