In allusion to the problem that the traditional algorithms which solve slant correction and segmentation have different effects on different types of character images, a new algorithm is presented. It improves the traditional algorithms by adding a new slant angle determination rule, by adjusting the threshold dynamically and postprocessing the regions. The new algorithm not only solves the problem that the traditional method is easily interfered by noise and complicated background and it has bad operability, but also realizes slant correction and segmentation in character images. The experimental results show that although the new algorithm’s numbers produced by the measurement of pixels are more than the algorithm of maximum between cluster variances, its segmentation effect is significantly better than the algorithm of maximum between cluster variances and threshold segmentation algorithm proposed by literature［7］, Meanwhile,the new algorithm is more robust than the traditional methods, and the running time meets the requirement of realtimeness completely.