欢迎访问《图学学报》
分享到：
摘要：传统的脉冲耦合神经网络(PCNN)在图像降噪时不能准确地定位噪声数据并去除
图像噪声。提出一种基于改进的PCNN 有效去噪方法。该方法在PCNN 模型上采用自适应的
突触连接系数，使之随不同神经元与其周围神经元相似程度的不同而自适应变化，提高噪声
数据的辨识度；同时将PCNN 神经元的点火频次记录在点火时间序列中，根据神经元点火次
数判断并滤出噪声点，实现更好地降噪效果。实验测试结果表明，该方法不仅可以准确地辨
识噪声数据，而且能够有效地去除图像的噪声点，具有较强的适应性和较好的边缘与细节保
护能力。
Abstract: Aiming at the problem that traditional pulse coupled neural network cannot be more#br# accurate positioning and reducing impulse noise in mage, an effective method for reducing impulsive#br# noise based on modified PCNN is presented. Adaptive synaptic connection coefficient is applied in#br# the PCNN model. To improve identification of the noise data, it is modified as variable value on the#br# similar degree between neurons and their surrounding neurons. Moreover the ignition time sequence#br# records firing frequencies of PCNN neurons, and noise points are identified and filtered according to#br# the ignition times. Thus this method achieves a better de-noising effect. Experimental results show#br# that the proposed method can not only identify noise data accurately, but also filter impulse noise#br# effectively. It has strong adaptability and good capability to protect edges and details of images.
