基于灰度迭代阈值脉冲耦合神经网络的图像分割

摘要

A new method, called Grayscale Iteration Threshold Pulse Coupled Neural Network ( GIT-PCNN), was proposed for image segmentation. The GIT-PCNN reduced the required parameters of conventional PCNN and the exponentially decaying threshold was improved to be related to the grayscale statistics of the original image. When GIT-PCNN was applied to image segmentation, no parameter or iteration time needs to be determined since the segmentation could be completed by one time of PCNN firing process. Therefore, GIT-PCNN did not require specific rule as the iteration stop condition. GIT-PCNN made good use of the grayscale information of the original image and the pulse characteristics of PCNN that the neurons associated with each group of spatially connected pixels with similar intensities tended to pulse together when partitioning images. The experimental results show that GIT-PCNN is better than classical PCNN-based segmentation algorithms on visual evaluation, subjective indices and speed performance.%为有效分割图像,提出了灰度迭代阈值脉冲耦合神经网络(GIT-PCNN).GIT-PCNN简化了传统PCNN模型,将其指数衰减的阈值改进为图像的灰度迭代阈值.GIT-PCNN分割图像时无需进行参数和循环次数选择,也无需使用特定原则确定循环结束条件,一次点火过程完成分割.GIT-PCNN分割图像时充分利用了图像的灰度信息和PCNN特有的空间邻近及像素灰度值相似集群发放脉冲提供的图像局部位置信息.实验结果表明,GIT-PCNN在主观及客观的分割性能和速度上均优于经典的PCNN分割方法.