Research On Super-resolution Reconstruction Algorithm Of Color Image Based On Sparse Regularization Model

With the rapid development of image processing and analysis technology,the demand for high resolution images and video information is increasing.Therefore,the research of super-resolution reconstruction for the reconstruction of ideal high resolution images from the actual low resolution images came into being.The technology of super resolution reconstruction does not change the existing physical imaging system.It only uses software technology to generate high-resolution images,which has great advantages in technology and cost.So it is more and more used in space exploration,high definition TV and medical image processing in recent years.Super-resolution reconstruction based on sparse representation model is a hot topic in the research of super-resolution reconstruction in recent years.In this paper,a single color image based on sparse representation is used to study the super-resolution reconstruction of the image.The main tasks include:1.Aiming at the problem of blurred image and poor visual effe ct in color image reconstruction,a super resolution reconstruction algorithm of color image based on L_2/3 sparse regularization model is proposed.The algorithm uses Gabor transform to extract the details of the color image.Instead of L ₀ sparse regular model,L_2/3 sparse regularization model is used to build super resolution reconstruction algorithm of color image based on sparse regular model,and making the algorithm more easy to solve,which not only reduces the computational complexity of the algorith m,but also effectively restores the detail information of the color image and eliminates the visual artifact of the reconstructed color image.2.A super resolution reconstruction algorithm for color images based on quaternion and sparse regularization model s is proposed for the missing correlation between channels in the super-resolution reconstruction algorithm based on sparse representation.The algorithm uses quaternion to represent each channel of the color image,and enhances the correlation between the channels.The L_1/2 sparse regular term is used instead of the L₁ sparse regular term to construct the super-resolution dictionary training and reconstruction model.In order to represent the details of the image more effectively,the high and low resoluti on training samples for reconstruction are constructed by means of the eliminating mean method.Experimental analysis shows that this method can not only remove color artifacts in reconstructed images,but also improve the quality of reconstructed images.