Font Size: a A A

Fast Super-resolution Reconstruction Based On Anisotropic Smoothing Terms

Posted on:2018-02-08Degree:MasterType:Thesis
Country:ChinaCandidate:S J ZhaoFull Text:PDF
GTID:2358330512476577Subject:Optical Engineering
Abstract/Summary:PDF Full Text Request
Image super-resolution reconstruction technology is designed to reconstruct high-resolution images with more abundant information by using the complementary information of low-resolution images of the same scene or the local structural features of single low-resolution images.Compared to the technical limitations of hardware methods and cost issues,through the computer software to enhance the resolution of the image has important practical significance.This paper introduces the BBF query method and bidirectional matching method to improve the efficiency of SIFT feature image registration.The SIFT feature is not affected by illumination,angle rotation and position translation,and can reflect the structural characteristics of the image.In the search of SIFT feature points,the exhaustive method is inefficient and the BBF search method can quickly locate the target feature points.In the process of feature point matching,bidirectional matching has higher accuracy than unidirectional matching.The experimental results show that the proposed algorithm has high registration precision.In this paper,a fast super-resolution reconstruction algorithm is proposed based on the traditional reconstruction iterative model.The key of the fast super-resolution reconstruction algorithm is to remove redundant processes in the reconstruction iterative model,which always occupy the majority of reconstruction computation.The image super-resolution reconstruction is divided into two parts.Firstly,a blurred high-resolution image is obtained by image registration.Then,the high-resolution image is reconstructed by iteration.This paper proposes a fast super-resolution reconstruction algorithm based on adaptive anisotropic smoothing term.Firstly,the Tikhonov regularization terms and the total variation regularization terms are analyzed,and a super-resolution reconstruction algorithm based on anisotropic smoothing terms is proposed.Anisotropic smoothing can be done to smooth the local structure to remove noise,while maintaining sharp edges between different local structures.In addition,this paper designs an adaptive parameter based on anisotropic smoothing term,which changes with the residual term and the regular term.Experimental results show that the proposed method not only reduces the time to reconstruct image,but also improves the quality.
Keywords/Search Tags:Super-resolution reconstruction, SIFT, Iterative model, Adaptive regularization parameter, Anisotropic smoothing term
PDF Full Text Request
Related items