光学与光电技术, 2019, 17 (4): 45, 网络出版: 2019-09-27  

压缩感知理论在太赫兹成像中的应用研究

Research of Compressive Sensing in Terahertz Imaging
邵哲明 1,*岳松 1,2左志高 1,2王友 1,2
作者单位
1 华中光电技术研究所-武汉光电国家研究中心, 湖北 武汉 430223
2 湖北久之洋红外系统股份有限公司, 湖北 武汉 430223
摘要
为了解决太赫兹成像系统中存在的成像速度慢、硬件复杂等问题, 利用压缩感知算法实现太赫兹成像系统性能的提升。通过对比离散小波基和离散傅里叶基对重构图像的影响, 发现离散小波基作为稀疏基时, 得到图像信噪比最优。然后验证了压缩采样匹配追踪、迭代加权二范数、贪婪基追踪等重构算法在太赫兹成像领域的应用可行性, 发现当测量比例大于40%时, 才能实现图像的重构。根据峰值信噪比、均方误差、算法运行时间等参数的对比结果, 发现迭代加权二范数算法相对空间匹配追踪算法, 峰值信噪比提升度达到20%, 均方误差下降度均大于50%。同时发现, 空间匹配追踪算法和子空间追踪算法的运行时间最少, 其中空间匹配追踪算法均小于0.5 s。为压缩感知理论在太赫兹成像领域的应用提供了新思路。
Abstract
In order to solve the problems of terahertz imaging system, such as slow speed and complicated hardware, the feasibility of compressive sensing theory to improve system performance of terahertz imaging is verified. By contrasting the imaging evaluation functions, it is found that the PSNR of the image is optimal, when the discrete wavelet basis is regarded as sparse basis. Then, the reconstruction algorithms, such as CoSaMP, IRLS, and GBP, are verified feasibility under algorithmically in terahertz imaging, and image reconstruction is realized when the measurement scale is 40%. According to the comparison of the peak signal to noise ratio, mean-square error, running time, it is found that the PSNR of IRLS is up to 20 percent, and the MSE of IRLS drop is more than 50 percent, comparing to the OMP. At the same time, running time of OMP is the least, which is smaller than 0.5 seconds.

邵哲明, 岳松, 左志高, 王友. 压缩感知理论在太赫兹成像中的应用研究[J]. 光学与光电技术, 2019, 17(4): 45. SHAO Zhe-ming, YUE Song, ZUO Zhi-gao, WANG You. Research of Compressive Sensing in Terahertz Imaging[J]. OPTICS & OPTOELECTRONIC TECHNOLOGY, 2019, 17(4): 45.

关于本站 Cookie 的使用提示

中国光学期刊网使用基于 cookie 的技术来更好地为您提供各项服务,点击此处了解我们的隐私策略。 如您需继续使用本网站,请您授权我们使用本地 cookie 来保存部分信息。
全站搜索
您最值得信赖的光电行业旗舰网络服务平台!