Noise reduction using genetic algorithm based PCNN method
Deng Zhang, Shingo Mabu, Kotaro Hirasawa · 2010
Pulse Coupled Neural Network (PCNN) has gained widely attention in image noise reduction as a non-linear filtering technique. Conventional PCNN-based methods usually are combined with median filter or step-by-step modifying algorithm. There are mainly two problems. In one hand, such filtering approaches blur the edge when smoothing an image. In the other hand, it is difficult to properly estimate the parameters of PCNN. Consequently, this paper presents an adaptive genetic algorithm based PCNN method (GA-PCNN) to restrain from additive white Gaussian noise (AWGN). Different from conventional PCNN-based methods, GA-PCNN utilizes an anisotropic diffusion filter to replace the median filter and optimizes the parameters of a simplified PCNN by means of adaptive genetic algorithm. Experimental results indicate that GA-PCNN has a better performance than the previous denoising techniques, i.e., median filter, Wiener filter, anisotropic diffusion filter, and the conventional PCNN based methods. Conclusions on the effectiveness of Gaussian noise reduction and edge preservation are carried out finally. Furthermore, the results will contribute de-noising in CMOS image sensors.