Image Denoising Method Based on ICA and BP Neural Network

Abstract

The image will inevitably be mixed with noise or interference signals in the process of acquisition and storage. For this reason, independent component analysis (ICA) and genetic Bayesian regularized BP neural networks are combined to deal with image denoising problems. Firstly, the image to be processed is separated into independent noisy images by ICA method. Then the noisy image is predicted by the genetic Bayesian regularized BP neural network to obtain a clear image. Experiments show this method can improve the PSNR and correlation coefficient of the image.