An Algorithm for Cell Microscopic Image Segmentation
The study of biological cell image segmentation is of great significance for accomplishing three-dimensional reconstruction of microstructure and ultrastructure of cell and it can also provide basis for disease diagnosis, and promote quantitative analysis of cell information, transfer of internal information of cell and research of cytometaplasia. In this paper, research is conducted mainly for edge extraction and adhesion segmentation of cell image and a cell microscopic image segmentation algorithm based on PCNN and Otsu is proposed. This algorithm applies Otsu to the selection of initial threshold of PCNN network. Otsu traverses the value of image pixels, seeks the maximum variance between the background and object and gets the optimal threshold. This threshold is optimal in statistical sense and it is used as the initial threshold of PCNN network. This paper reveals the essential relationship between pulse coupled neural network (PCNN) and mathematical morphology, namely that the parallel transmission of pulse of PCNN is totally equal to the erosion operation of certain structural elements in mathematical morphology. This paper makes use of this characteristic, revises the parameters, and realizes two-way transmission of PCNN to conduct erosion and dilation simultaneously. In this way, it can quickly fill small cavity and optimize the segmentation result. The simulation experiment proves that the algorithm of this paper can lead to a relatively complete result of tissue cell image segmentation.