光学 精密工程, 2013, 21 (6): 1576, 网络出版: 2013-07-01
采用非局部均值的超分辨率重构
Super-resolution reconstruction using nonlocal means
非局部均值 超分辨率重构 运动估计 滤波器 去噪 nonlocal means super-resolution reconstruction motion estimation filter denoising
摘要
由于传统的超分辨率重构无法在工程应用中对含有局部运动图像进行有效的运动估计及重构,本文提出一种采用非局部均值(NLM)的超分辨率重构方案。简要介绍了具有较好去噪特性的非局部均值滤波器,分析了超分辨率重构的代价函数,根据构造出的非局部均值超分辨率重构算法的代价函数及其求解,对提出的方案进行进一步的优化和化简,最后得到一种易于工程实现的重构算法。实验结果表明,提出的算法不仅具有NLM算法的优点,即不需进行显式的运动估计就能得到更清晰、细节更丰富的重构图像;而且重构速度比简化前的NLM算法提高将近30%,有望应用于具有复杂运动的图像的超分辨率重构。
Abstract
As conventional super-resolution algorithms can not implement the motion estimation and reconstruction for an image with local motion in practical engineering applications, this paper proposed a super-resolution reconstruction algorithm based on NonLocal Means (NLM). First, the NLM filter, one of the successful denoising filters in recent years, was introduced briefly. Then, the details concerning its application to super-resolution were analyzed by creating a super-resolution cost function. By considering the practical situations and the need of the engineering facet, a scheme to simplify the procedure in the NLM super-resolution algorithm was proposed. The experiment results show that the simplified algorithm can not only effectively implement super-resolution reconstruction to get a clear and detailed image without explicit motion estimation, but also can obtain a reconstructed speed higher 30% than that of conventional algorithms. It can satisfy the practical needs of engineering set-tings and is expected to reconstruct the high resolution image with complex motion.
李家德, 张叶, 贾平. 采用非局部均值的超分辨率重构[J]. 光学 精密工程, 2013, 21(6): 1576. LI Jia-de, ZHANG Ye, JIA Ping. Super-resolution reconstruction using nonlocal means[J]. Optics and Precision Engineering, 2013, 21(6): 1576.