Adaptive blocks for skeleton segmentation in handwritten Thai character

The structural analysis in character recognition systems is mostly dependent on stroke segmentation process. To improve the system, a novel segmentation algorithm for handwritten Thai characters, adaptive block, is proposed. A character is segmented into basic elements by using adaptive block. The block can be formed by using the distinctive features of Thai characters such as cavity, local minimum and maximum point. After segmentation process, each segment is coded as primitive codes. Experiments were performed on a lightly constrained handwritten Thai character data set that was obtained from ten Thai peoples. The experimental results have shown the proposed method is robust against the diversity of handwritten Thai character and its codes are satisfactory to recognize