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 denoisi...

Full description

Saved in:

Bibliographic Details
Published in:	2015 IEEE International Conference on Information and Automation pp. 2720 - 2724
Main Authors:	Xiufen Ye, Lin Wang, Huiming Xing, Le Huang
Format:	Conference Proceeding
Language:	English
Published:	IEEE 01.08.2015
Subjects:	deep learning Image denoising Neural networks Noise Noise reduction Speckle stacked sparse denoising autoencoder Testing Training
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!

Description
Summary:	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.
DOI:	10.1109/ICInfA.2015.7279746