基于差值控制细胞神经网络图像滤波器

差值控制细胞神经网能够实现灰度图像滤波等复杂运算.针对原有差值控制细胞神经网中值滤波器在稳定性和可实现性上存在的不足,提出了一种伪中值滤波器(CNN PM-filter),进而引入Mask图构造了选点式伪中值滤波器.从实验结果和相关度分析可以看出,本文提出的两种滤波器在改善稳定性与实现性的同时,没有影响到滤波器的性能,而选点式伪中值滤波器能有效降低滤波造成的模糊图像,取得更佳处理效果.

Image filters based on difference-controlled cellular neural networks

Difference-controlled Cellular Neural Networks (CNN) can realize some complex operations more convenient than standard CNN. To improve the stability and realizability of existent median filters based on difference-controlled CNN, a pseudo median filter based on difference-controlled CNN (CNN PM-filter) was presented. In order to reduce image blur caused by filtering, a selective CNN PM-filter was studied too. The results of signal/ noise ratio and correlation degree show that the stability and realizability of the two filters in paper were improved. The CNN PM-filter's performance is a little better than a standard median filter; the selective CNN PM-filter can suppress impul noise and simultaneously reduce image blur effectively.