A multiphase chan-vese model using edge information and bias field correction for image segmentation

Intensity inhomogeneity and the curve of level set evolution misguiding the edge of target cause considerable difficulty in image segmentation. This paper proposes a new method in which a multiplicative bias field function is converted into its linear form with logarithmic transformation; the intensity information in local region is extracted to the bias field function by using kernel function; a new bias field correction model is built based on variational method. In addition, the edge indicator function is rebuilt by comprehensive utilization of the direction information of the level set function and image; this improves the ability of the evolving curve to recognize target boundaries. Experimenntal results for synthetic and real images and their analysis show preliminarily that the performances of our method are desirable.