一种基于PCNN的图像分割改进算法

An Improved Image Segmentation Algorithm Based on PCNN

为了实现脉冲耦合神经网络(PCNN)在图像分割过程中模型参数与迭代次数的少量化,提出了一种改进的PCNN快速图像分割算法。算法对PCNN模型进行了简化,将传统PCNN模型中恒定的连接系数与神经元所在像素点的像素值联系起来,去除了PCNN分割图像过程中的人工设置参数过程,并根据图像灰度统计特性,将动态阈值转变为恒定阈值,仅一次迭代便可完成图像分割。实验结果表明:算法的分割结果主观视觉感受良好,时间复杂度低,优于对比算法。

In order to reduce the number of model parameters and iteration times of the Pulse Coupled Neural Network(PCNN)in image segmentation,an improved PCNN fast image segmentation algorithm is proposed.PCNN is simplified and the constant connection coefficient is linked with pixel grayscale value of corresponding neurons in the algorithm.The algorithm does not need to set PCNN parameters artificially in image segmentation and the dynamic threshold is transformed into constant threshold according to grayscale statistical characteristics of the image so that only one iteration time is needed for image segmentation.The experimental results show that the proposed algorithm has good performance of subjective visual perception and lower time complexity,and it is better than comparison algorithm.