QQ客服 官方微博 反馈留言

题名	结构压缩感知在多光谱强度关联成像中的应用研究
作者	谭诗语
学位类别	硕士
答辩日期	2015
授予单位	中国科学院上海光学精密机械研究所
导师	韩申生
关键词	图像重建 压缩感知 多光谱图像 先验图像 奇异值分解
其他题名	The Application Research of Structured Compressed Sensing in Multispectral Ghost Imaging
中文摘要	多光谱图像是对同一物体进行多谱段成像所得到的二维图像组，其包含多个连续的光谱信息，数据量大、计算复杂度高，对图像的重建提出了很大的挑战。为了提高多光谱图像的重建质量，本文结合压缩感知理论，利用多光谱图像的空间相关性和谱间相关性，提出了基于先验图像约束的多光谱压缩感知（PICHCS）重建方法、基于有效奇异值约束的多光谱压缩感知（ESVHCS）重建方法以及将PICHCS和ESVHCS相结合的EPHCS算法。PICHCS算法将相邻谱段的初始图像取平均获得高信噪比的先验图像。先验图像与目标图像相减可以使优化目标稀疏化，并使得重建结果具有与先验图像类似的高信噪比特性。ESVHCS算法基于有效秩准则和有效奇异值概念，将有效奇异值的和作为谱间约束加入到目标函数中，能减小重建图像的噪声，提高重建质量。文中通过数值模拟和实验验证了这些重建算法的可行性，并在不同的采样率、信噪比条件下和全变差低秩（TV_RANK）联合重建算法进行了对比研究。结果表明，PICHCS、ESVHCS和EPHCS算法可以有效提高多光谱图像的重建质量，从而降低对数据采样率和系统信噪比的要求。 本文的内容主要包括： 1、介绍了压缩感知理论及其算法，说明了压缩感知理论的研究意义以及发展趋势 2、对多光谱成像技术和关联成像技术做了简单的介绍，说明了光谱成像技术目前遇到的瓶颈以及发展趋势，并介绍了多光谱压缩感知成像技术的发展历史和研究现状。 3、介绍了一种基于相位调制的准单色光单次曝光压缩感知关联成像技术，并在其基础上介绍了一种基于相位调制的单次曝光多光谱压缩感知关联成像技术. 4、分别介绍了PICHCS算法、ESVHCS算法和EPHCS算法的理论模型和实现方法，通过数值模拟和实验验证了这些重建算法的可行性，并在不同的采样率、信噪比条件下和全变差低秩（TV_RANK）联合重建算法进行了对比研究。
英文摘要	Multispectral images are a group of two-dimensional images of the same object which contain a number of successive spectral information. The large data volume and high complexity of the multispectral images presents a great challenge of image reconstruction. Both space correlation and spectral correlation of multispectral image are utilized in order to improve the quality of reconstruction image. According to compressed sensing theory, we proposed PICHCS algorithm based on prior image constraints, ESVHCS algorithm based on the effective singular value constraints and EPHCS algorithm which combines PICHCS and ESVHCS algorithm. In PICHCS algorithm, prior images are obtained by averaging the adjacent spectral primitive images. By subtracting each target image with the corresponding prior image, the obtained difference images can be expected to be sparse and by reconstructing the difference images, some of the high SNR characteristics of the prior image may be retained in the reconstructions. ESVHCS algorithm is based on the concept of effective rank criterion and effective singular value. The sum of effective singular values is added to the objective function as spectral constraints, which can reduce the reconstructed image noise and improve the reconstruction quality. The feasibility of the method is verified by numerical simulations and experiments. Comparative studies are made for reconstructions obtained with the total variation and low rank (TV_RANK) joint algorithm and those with PICHCS under different sampling ratios and SNRs. The results indicate that PICHCS, ESVHCS and EPHCS can improve the reconstruction quality of hyperspectral images and reduce the required number of measurements. This paper mainly consists of these content: Firstly, compressed sensing theory and its algorithm are introduced. Then, we state the Significance and developing trend of compressed sensing . Secondly, a brief introduce of multispectral imaging and ghost imaging is presented. The history and developing trend of compressive sensing multispectral imaging are also introduced. Thirdly, a quasi-monochromatic ghost imaging technology with passive thermal light by phase modulation and a multi-spectral ghost imaging technology with passive thermal light by phase modulation are introduced. Lastly, We introduced the theoretical model and implementation methods of PICHCS algorithm, ESVHCS algorithm and EPHCS algorithm. The feasibility of the method is verified by numerical simulations and experiments. Comparative studies are made for reconstructions obtained with the total variation and low rank (TV_RANK) joint algorithm and those with PICHCS under different sampling ratios and SNRs.
语种	中文
内容类型	学位论文
源URL	[http://ir.siom.ac.cn/handle/181231/16942]
专题	上海光学精密机械研究所_学位论文
推荐引用方式 GB/T 7714	谭诗语. 结构压缩感知在多光谱强度关联成像中的应用研究[D]. 中国科学院上海光学精密机械研究所. 2015.

个性服务

查看访问统计

相关权益政策

暂无数据

收藏/分享

除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。