The most significant parts in Figure 4 are the image binarization, digital features selection and strokes confirmation. Image binarization. As the background of the image isusually quite complicated, choosing a single threshold for image binarization often introduces interferences which could cause a failed binarization. So the zoning binarization method is used to avoid such situation [8, 9]. This method is to cut the image into many pieces so that each piece would enjoy a monotonous background and then their thresholds will be calculated respectively. According to the thresholds, these pieces are binarized and combined sequentially. An example of image binarization is shown in Figure 5. As we see, although suffering severe optical reflection, the original image is well binarized.