基于刃边法的序列图像盲超分辨率重建算法  

The Image Sequence Blind Super-resolution Algorithm Based on Slanted-edge Method

作　　者：张晓林[1] 何小海[1] 李滔[1] 梁子飞[1] 

机构地区：[1]四川大学电子信息学院,四川成都610065

出　　处：《四川大学学报（工程科学版）》2013年第S1期84-89,共6页Journal of Sichuan University (Engineering Science Edition)

基　　金：国家自然科学基金资助项目(61071161)

摘　　要：为了减少点扩展函数(PSF)估计误差对盲超分辨率重建结果的影响,提出了1种新的序列图像盲超分辨率重建算法。首先采用刃边法实现对成像系统点扩展函数的准确估计,然后将估计的点扩展函数引入迭代反投影超分辨率重建算法中,进行高分辨率图像的重建。实验结果表明,本文算法运算时间较短,不受限于低分辨率图像的模糊程度,重建结果不但具有较高的峰值信噪比(PSNR)和结构相似度(SSIM),而且真实分辨率的提升幅度更大。A new image sequences blind super-resolution algorithm was proposed to reduce the influence on super-resolution construction caused by point spread function estimation error.The slanted-edge method was used to achieve an accurate estimate of the image system's point spread function,which was then combined the point spread function with the iterative back projection super-resolution construction algorithm to finish the super-resolution construction.Experimental results indicated that the proposed method needs shorter operation time and doesn't limited by the blur degree of the low resolution image.The construction result is of higher peak signal to noise ratio(PSNR) and structure similarity(SSIM) and lager extent of the image's true resolution.

关 键 词：盲超分辨率重建 刃边法 迭代反投影 序列图像 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]