A Vehicle License Character Segmentation Algorithm Based on Local Chan-Vese Model

In this study, an improved CV model (MCV model) was proposed on the concept of using one level set function for each region; it needs less number of iterations and improves the efficiency of image segmentation contrast to CV model. But it was easily trapped in local optima for the influence of initial contours of regions; also it often led to poor segmentation results for images with intensity inhomogeneity. To overcome these problems, this study used the k-means algorithm to pre-segment the image to get the initial curves, and then added the local information to the total energy function. Finally, a number of relevant segmentation results are demonstrated that the new model can get better results in an efficient way for multiphase images.