Fast compressed sensing analysis for super-resolution imaging using L1-homotopy
Fast compressed sensing analysis for super-resolution imaging using L1-homotopy
复制标题
DOI：
10.1364/oe.21.028583
复制
发表时间：
2013-11-18
期刊：
影响因子：
3.8
通讯作者：
Zhuang, Xiaowei
中科院分区：
文献类型：
--
作者：
Babcock, Hazen P.;Moffitt, Jeffrey R.;Zhuang, Xiaowei
在基于单分子切换和定位的超分辨率成像技术中，获取超分辨率图像的时间受到每个成像帧可以精确定位的荧光发射体的最大密度的限制。为了提高成像速率，最近已经开发了几种方法来分析具有更高发射体密度的图像。一种强大的方法使用基于压缩感知的方法，与其他报告的方法相比，将每成像帧的可分析发射器密度增加数倍。然而，这种方法的计算成本，它使用内部点的方法，是高的，和一个典型的40亩米× 40亩米的视野超分辨率电影的分析需要在高端台式个人电脑上的数千小时。在这里，我们展示了一种替代的压缩感知算法，L1同伦（L1H），它可以生成超分辨率图像重建，这些重建与使用内点方法获得的重建基本相同，根据发射体密度，时间减少一到两个数量级。此外，对于具有不同发射体密度的实验数据集，L1H分析比内点方法快300倍。这种计算时间的大幅减少应该允许压缩感知方法被常规地应用于超分辨率图像分析。(C)2013年美国光学学会
In super-resolution imaging techniques based on single-molecule switching and localization, the time to acquire a super-resolution image is limited by the maximum density of fluorescent emitters that can be accurately localized per imaging frame. In order to increase the imaging rate, several methods have been recently developed to analyze images with higher emitter densities. One powerful approach uses methods based on compressed sensing to increase the analyzable emitter density per imaging frame by several-fold compared to other reported approaches. However, the computational cost of this approach, which uses interior point methods, is high, and analysis of a typical 40 mu m x 40 mu m field-of-view super-resolution movie requires thousands of hours on a high-end desktop personal computer. Here, we demonstrate an alternative compressed-sensing algorithm, L1-Homotopy (L1H), which can generate super-resolution image reconstructions that are essentially identical to those derived using interior point methods in one to two orders of magnitude less time depending on the emitter density. Moreover, for an experimental data set with varying emitter density, L1H analysis is similar to 300-fold faster than interior point methods. This drastic reduction in computational time should allow the compressed sensing approach to be routinely applied to super-resolution image analysis. (C) 2013 Optical Society of America