This paper explores new approaches to recognize Chinese characters in images through touch sense. We extract the gradient image fused with brightness and chrominance information by the flow-based Gauss difference algorithm, and use the edge tangential flow method to smooth image contour, to obtain smooth character contours with significant edge features. Then, we use a tactile generation algorithm with five mechanical elements to generate stable and continuous tactile information based on the normal vector of triangular patches of the surface. To verify the feasibility and effectiveness of the proposed methods, we developed a Chinese character tactile sensing system. Experiments show that Chinese characters in images can be recognized by touch sense with higher recognition accuracy rate.