Blood Cell Segmentation Based on Improved Pulse Coupled Neural Network and Fuzzy Entropy

In the field of biomedical image processing, because of the low intensity and brightness of the cell image, and the complex structure of the cell image, the segmentation of cell images is very difficult. A large number of studies have shown that the Pulse Coupled Neural Networks (PCNN) is suitable for image segmentation. However, the traditional PCNN must set a large number of parameters in image segmentation, and the optimal number of iterations cannot be automatically determined. In this paper, a new improved PCNN model is proposed. The work of improved PCNN includes the acceptance portion of the PCNN model being simplified and the connection portion of PCNN being improved. In addition, the maximum fuzzy entropy is used as the criterion to determine the optimal number of iterations. Experimental results on blood cell image segmentation show that this proposed method can automatically determine the number of loop iterations and automatically select the best threshold. It also has the characteristics of fast convergence, high accuracy and good segmentation effect in blood cell image segmentation processing.