Super Resolution Reconstruction Algorithm For Gray Image Based On Sparse Representation

With the rapid development of science and technology,today's society has entered an information age,Image has been widely used in the communication and transmission of information.Image has become one of the most commonly used information carriers in human social activities.How to effectively improve the resolution of images and improve the quality of images has become an urgent need.Image super-resolution reconstruction technology has become an important topic in the field of visual image processing.In recent years,sparse representation and compression sensing techniques have been continuously improved,which makes the image super-resolution reconstruction algorithm based on sparse representation develop rapidly.In this paper,the research status of image super-resolution reconstruction algorithm is described in detail,and the theoretical basis of the algorithm is summarized.This paper focuses on the learning based super-resolution reconstruction algorithm for gray-scale images.The main work is as follows:1.In order to solve the problem that the general super-resolution reconstruction method is not good enough for gray image reconstruction,it is difficult to recognize it.In this paper,a super-resolution reconstruction algorithm for gray image is proposed.In order to make the training dictionary more reasonable and targeted,the gray-scale image is classified and processed,and a specific dictionary is trained for each kind of gray-scale image.The fusion clustering algorithm trains dictionaries to make the generated dictionaries more suitable for different types of gray-scale images.2.At present,the solution of L1 regular model is not sparse enough,and the precision of sparse representation coefficient can be further improved.In this paper,the super-resolution reconstruction model is constructed by using L1/2 norm instead of L1 norm.The semi-threshold iterative algorithm combined with the accelerated approximate gradient algorithm is used to solve the L1/2 regularization problem,and a super-resolution reconstruction model is established.The L1/2 regularization theory is applied to dictionary learning.The experimental results show that the proposed algorithm is superior to the L1 regularization super-resolution reconstruction algorithm.