基于脉冲耦合神经网络的图像分割

针对传统脉冲耦合神经网络(PCNN)模型在图像分割时需要设置较多参数和不能准确分割低对比度图像的问题,提出一种简化的PCNN模型和改进算法。在简化模型中减少了在传统PCNN模型中需要设置的参数的数量；在改进算法中根据图像像素空间和灰度特征自适应设置模型参数,并根据图像灰度直方图求出灰度期望均值作为图像分割阈值,因此该算法无需选择 循环迭代次数,只需一次点火过程就能实现图像的有效分割。实验结果表明,该方法能准确分割图像,纹理细节清晰,分割结果优于人工调整参数的PCNN方法和Otsu方法。

Image Segmentation Based on Pulse Coupled Neural Network

Traditional pulse coupled neural network (PCNN) model needs to set a lot of parameters in processing image segmentation and can not segment the images with low contrast precisely.In order to solve the problems,an improved image segmentation algorithm was proposed based on a simplified PCNN model.In the simplified mode,the number of parameters required in the traditional PCNN model was reduced.In the improved algorithm,the model parameters were set adaptively according to the image pixel space and gray features,and the image grayscale expected mean was obtained as the image segmentation threshold according to the image gray histogram.Therefore,the improved algorithm has no iteration stop condition need to choose,just once the ignition process the method can complete the image segmentation effectively.The experimental results show that this method is accurate in image segmentation,especially in image texture details,and the final result is better than some methods,such as manual adjustment method of PCNN parameters and Otsu method.