Infrared image denoising based on convolutional neural network
Cheng Jian Sun, Mingqiang Pan, Bin Zhou, Zong-jian Zhu · 2018
In order to remove the noise in the infrared image more effectively, a denoising method based on convolutional neural network is proposed. This method constructs a network structure composed of convolutional subnet and deconvoluted subnet, the convolution subnet extracts the features of the image, and the deconvolution subnet reconstructs the original image through the feature map. The image with noise is obtained by adding noises to the image without noise to form a training set and a test set, Tensorflow is used to train infrared image samples in training set to fit the mapping function between image with noise and image without noise, and use the learned mapping function to denoise the test set. Experimental results show that compared with the traditional denoising method, the proposed method can effectively remove the noise in the infrared image under high noise environment, and achieve a higher peak signal-to-noise ratio.