Image Super-resolution Reconstruction Based On Residual Dictionary Learning

Super-resolution image reconstruction refers to the restoration of low-resolution images by digital image processing without changing the device of image acquisition and the external environment,and obtaining high-resolution images which are close to the original image in visual effects..The traditional image reconstruction algorithm is simple to achieve but the final effect is not so well.In this paper,according to the theory of machine learning,a super-resolution image reconstruction algorithm based on residual dictionary is proposed,in which not only the visual effect of reconstructed image is improved,but also the objective evaluation index peak,signal-to-noise ratio(PSNR)and structural similarity(SSIM),is increased.The specific work of this paper is as follows:1.An image reconstruction algorithm combining residual dictionary with ordinary dictionary is proposed.The SVR(Support Vector Regression)is used to train the sample to generate a learning model called a dictionary.Then image reconstructed by the dictionary is compared with the original high-resolution image to obtain the residual image.Finally the residual image and the reconstructed image are trained again by SVR to obtain a residual dictionary.In the prediction stage,the two dictionaries are used to reconstruct the image in order,which can better restore high-frequency information such as edges of the image.So a high-quality reconstructed image is obtained.2.The effectiveness of the proposed algorithm is verified.The algorithm of this paper is compared with the traditional bicubic interpolation,convex set projection and convolutional neural network from the visual effects and objective evaluation.The experimental results show that the performance of this method is improved in both aspects.3.The scope of application in the algorithm is studied.Five kinds of images such as landscape,face and vehicle were selected,and the algorithm was used to carry out experiments and the quality of reconstructed images was analyzed.The experimental results show that the proposed algorithm can achieve better reconstruction effects when dealing with face and vehicle images.When processing aerial images,the reconstruction effect is not ideal.