Image super-resolution reconstruction method and system

Abstract

The invention provides an image super-resolution reconstruction method and system. According to the method, training samples are clustered by using a student t distribution hybrid model, then a reconstruction projection matrix of each type of training samples is constructed offline by using ridge regression, and finally super-resolution reconstruction is performed by using the most suitable projection matrix, so that the quality of image super-resolution reconstruction can be improved on the premise of ensuring the reconstruction efficiency.

Description

technical field [0001] The invention relates to image processing, in particular to a method and system for image super-resolution reconstruction. Background technique [0002] Images can accurately and intuitively reflect information in nature. Image quality depends largely on the resolution of the image. High-resolution images have richer details and contain more information, which not only has better visual effects, but also facilitates subsequent processing of image processing programs. [0003] For example, in the field of video security, it is often necessary to save surveillance videos for a long period of time. Due to the limitation of storage capacity, the resolution of videos captured by surveillance cameras is usually relatively low. When an abnormal situation occurs, low-resolution images often cannot provide effective information. The use of super-resolution technology can improve the image resolution, which is convenient for relevant personnel to deal with abn...

AI Technical Summary

Problems solved by technology

(2. Yang J, Wright J, Huang T S, et al. Image super-resolution via sparse representation [J]. IEEE transactions on image processing, 2010, 19(11): 2861-2873 .) Compared with the interpolation algorithm, the image super-resolution reconstruction algorithm based on sparse representation has greatly improved the reconstruction results, but sparse decomposition is required in the reconstruction process, which affects the operating efficiency, and only a single global dictionary is trained Yes, it limits the ability of the model to reconstruct images of different structures

(4. Zhang Y, Li K, Li K, et al.Image super-resolution using very deep residual channel attention networks (image super-resolution research using very deep residual channel attention network)[C].Proceedings of the European Conference on Computer Vision.2018:286-301.) Although the reconstruction results of the image super-resolution reconstruction algorithm based on the convolutional neural network are very good, the network scale with millions of parameters is too large, and the reconstruction requires a lot of computing resources , which limits its practical application

Images