基于局部方向正则化的快速图像插值算法  

A fast interpolation algorithm based on local direction regularization for image resolution

作　　者：孙冬[1] 李萌 高清维[1] 卢一相[1] 竺德 SUN Dong;LI Meng;GAO Qingwei;LU Yixiang;ZHU De(School of Electrical Engineering and Automation,Anhui University,Hefei 230601,China;Network Information Center,Anhui University,Hefei 230601,China)

机构地区：[1]安徽大学电气工程与自动化学院,安徽合肥230601 [2]安徽大学网络与信息中心,安徽合肥230601

出　　处：《安徽大学学报（自然科学版）》2020年第6期28-35,共8页Journal of Anhui University(Natural Science Edition)

基　　金：国家自然科学基金资助项目(61502007);安徽省自然科学基金资助项目(2008085MF192,2008085MF183);安徽省教育厅自然科学重点项目(KJ2018A0012,KJ2019A0023,KJ2019A0022);赛尔网络下一代互联网技术创新项目(NGII20180612,NGII20180312,NGII20180624)。

摘　　要：自然图像中丰富的结构化特征反映了信号中广泛存在的时空冗余性,该特性在去噪、去模糊、修复、插值和超分辨率重建中具有广泛的应用.考虑到人类视觉系统对图像边缘结构的敏感性,提出了一种基于局部方向正则化的快速图像插值算法.该算法根据图像的局部方向图,通过构造符合目标图像先验模型的正则化方程,从双立方预插图像为初始点,以迭代的方式实现复原方程的求解,对低分辨率图像中的边缘结构进行拟合与增强,实现对高分辨率目标图像的重建.该方案原理简单,完全在空域中实现,不涉及复杂优化方程的求解,计算复杂度低,易于实现.实验表明,所提方案可以有效地增强图像的分辨率,能够达到提升清晰度的目的.Natural images are highly structured,which reflects strong time-spatial redundancy and self-similarity underlying pixels,performs a key role in solving inverse problems such as image denoising,deblurring,inpainting,interpolation and super-resolution.By considering the fact that the human vision system is sensitive to image edge structure,a fast image interpolation algorithm based on local direction regularization was proposed in this paper.According to the local direction map of the image,the algorithm established the regularization equation that conformed to the prior model of the target image,and takes the bicubic interpolation version as the initial point to iteratively solve the restoration equation,where the edge structures and contours in low resolution input was reconstructed to estimate the target high resolution output.The appeal of the proposed method was its relative simple interpolation scheme,without solving complex optimization equations and was easy to use.Experimental results showed that the proposed method could effectively enhance the resolution and clarity of image,with satisfying subject visual quality and PSNR(peak signal to noise radio)value.

关 键 词：方向正则性 分辨率增强 图像滤波 插值 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]