Flaw Defect Enhancement Using Image Processing Approach Of The Fractional Differential Operator

In the past 20 years Fabric defect detection as a branch of automatic detections have received a number of scholars to research. There are many types of the fabric defect,such as the lose of latitude line,hole,knot and so on.Due to the many types of defects and the little difference of the defects area and the normal area, the fabric defect detection becomes very difficult.In the past fabric defect detection is usually done manually. As the person's energy is limited and costs are high, so how to use computers for fabric defect detection received at-tention, including home and abroad scholars.They have made a number of theoretical results and some actual detection systems.In accordance with the detection time, fabric defect detection can be divided into two categories, one category is an online defect detection, another kind of off-line defect detec-tion. Online defect detection mainly detects in the weaving process of detection, so that when flaws were found in the process of weaving, you can stop the machine timely and connect the line. The online test will give textile manufacturers to bring huge profits.But because of this need for fast real-time detection algorithms, the online detection can only be carried out under the special defects at present. Off-line defect detection is relatively easier than the online detection.The main purpose of fabric defect detection is to recognise defective area from the normal texture regions. In general, the normal fabrics with a complex texture is a certain periodicity in space. The appearance of defects breaks the frequency of the original, and this is the base that we are able to separate flaws in the region. However, an image with defect area has smaller in the area of defects, and the gap on the gray value is not great. In this paper,the object is the greige cloth with the absence of the longitude line.We collected the actual fabric which is small in the gray range and the defects area is a narrow-band.Adopted by a large number of experiments and theoretical analysis, we found that the classic image processing methods can not play significant results. This is mainly because the gray of region having flaws did not change significantly and can not achieve to enhance the defect region. Through a lot of analysis, we believe fractional differential operator image processing method is suitable for our fabric image. So we introduced the fractional differen-tial operator to carry out the fabric image filtering, and achieved some good results.For the fractional order research can be traced back to the 17th century. Fractional differentiation is the promotion of the mathematical integer-order differential. Fractional differentiation intrinsically linked to the fractal, Fractional differentiation is one of the foun-dations of fractal mathematics. Neighborhood of digital image between pixel and pixel gray value have the high degree of similarity. The high degree of similarity of fractal information usually represent by the complex texture. The science have confirmed that the fractional and fractal dimension of the mathematical method is the best man's description of natural phenomena. For this reason that we introduce fractional differential operator for this fabric image with complex texture.As well known that integer-order differential operator has enhanced the image, such as Sobel first-order differential operator, Laplace second-order differential operator. This opera-tor generally satisfied:(1)In the flat segment (gray constant region) differential value is zero; (2) In the gray-scale step or the starting point of slope differential value is non-zero; (3) along the slopes the differential t value is non-zero. The operators in highlighting the details of the image, or to enhance the details can get quite work. However,because the flaws regions have only a very small area, the direct use of these integer-order differential operator may not achieve the purpose of enhancing defect region. For our fabric images, in the experiment this thing happened. Through the theoretical study of the fractional differential operator, we found that fractional differential operator can not only enhance information of the mutation in its texture details(the image of high-frequency components), but also retains region of the normal texture with the non-linear way. This is also the the biggest difference between the fractional and integer-order differential operator. So we introduced the fractional differential operator in our fabric images. Using the fractional differential operator we can constructe mask in the eight directions on the fabric images, respectively, and then we can compute the eight gray values in the eight direction.We used the maximum value of the eight num-bers as our gray value in the new image.In our experiment images can greatly enhance the defect-oriented regional.The fabric image can inevitably bring some noise in the process of collecting images. In this paper,through a large number of experiments and theoretical analysis we found that Gauss and Gabor filtering for smoothing effect of the fabric image can denoise obviously. The fractional differential operator in some of the fabric image gets some good results.