Study On The Image Super Resolution Based On The Auto Encoder

Image super resolution is the process of obtaining high resolution images from one or multiple frame low resolution images using image processing techniques.It is a kind of method which is not limited by the hardware equipment,and can obtain the high resolution image more economically.Image super resolution technology is widely used in computer vision and image processing,such as medical diagnosis,video surveillance,text recognition and so on.In this paper,we study the image super resolution based on learning by using the autoencoder.The main work is as follows:Firstly,an image super-resolution method based on coupled autoencoder integrated the image edge information is proposed.In this method,the edge information of the image and the gray level information are taken as inputs,and the intrinsic features of the low resolution image blocks and the high resolution image blocks are studied by using the autoencoder.Then a single-layer neural network is used to establish the nonlinear mapping relationship between the low resolution image blocks and the intrinsic features of the high resolution image blocks.The experimental results show that the high resolution image obtained by the proposed method has some improvement in both numerical and visual effects.Secondly,an image super-resolution method based on sparse autoencoder is proposed.On the basis of the previous work,the sparsity of the hidden layer neurons of the autoencoder is added to improve the validity of the intrinsic representation.At the same time,the reconstructed high resolution image is iteratively adjusted and the approximate image is processed in order to further improve the reconstruction quality.The experimental results show that the reconstructed image has good effect both in visual effect and numerical results.