Denoising Hybrid Noises in Image with Stacked Autoencoder

A method based on sparse denoising autoencoder for denoising hybrid noises in image is proposed in this paper. The method is experimented on natural images and the performance is evaluated in terms of peak signal to noise ratio (PSNR). By specifically designing the training process of sparse denoising autoencoder, our model not only achieves good performance on single kind of noises, but also is relatively robust to mixed noises, which are more widely existed in practical situation. Autoencoder is a major branch of deep learning. It has been used in many applications as the method to exact features for its ability to represent the input data. Applying autoencoder to image denoising has been achieved good performance. Further research was deployed to find that autoencoder method is relatively robust compared with BM3D. And a sparse denoising autoencoder model is employed to train the network and it works well for the hybrid noise situation.