Noise classification of ICF images using a convolutional neural network (CNN)

In this paper, noise classification of ICF images is performed. One hundred thousand synthetic ICF images are generated to classify different kinds of noise. The noise of ICF images is neither additive nor does it follow any typical distributions. So, a deep neural network model, CNN, is designed to classify ten different noise distributions. Synthetic images are used as input to the model. With two hidden layers and an Adam optimizer, 91% accuracy is obtained. Experimental images are tested with the saved model, and the result is shown in this paper. Further study is needed to improve the accuracy of the model.