Level set method of cell image segmentation based on combinations of edge, region and prior information

Inhomogeneity of intensity often appear in the cell images and hence image segmentation may face lots of problems. For the purpose of solving the problems brought about by cell images with inhomogeneity intensity, low contrast and edge blurring, this paper presents a new level set method. Our model combines an external energy function with a regular term. The former combines the prior information with edge gradient and region information of cell images, and the latter makes our function approach to a distance function with a sign. Due to this internal energy, our model completely doesn't need to reinitialize. Experiments on mammary cell images show that our model has good segmentation performance on mammary cell images which are superior to watershed and clustering, DRLSE, Yang et al. segmentation methods.