基于压缩感知理论的图像重构技术

通过CS理论在工程应用上典型的正交匹配追踪算法,实现了一维信号和二维图像的精确重构,针对该方法中出现对整幅图像进行采样计算时需要大量的观测矩阵存储空间,并且重构过程中耗费了大量的时间等问题,提出了一种基于OMP算法的改进方案,将图像进行分块压缩感知。通过实验分析,以上问题得到了解决,重构图像的质量在没有增加计算复杂度的前提下也得到了提高。     

基于感知域测量值的自适应压缩采样方法

现有的自适应采样方法需要在采集端获得原始信号,这在实际的蒸发波导相关气象要素压缩感知过程中不易实现。针对这一问题,提出了一种以测量值数据特征为依据的采样率自适应设定方法。该方法以能量表征各信号块的重要性,以观测值能量近似估计原信号块能量,根据感知域中测量值的能量变化自适应地为各块设定采样数目。理论分析和实验表明,该方法的重构性能优于传统自适应方法,相同采样率下可获得更高的重构信噪比。     

压缩感知测量方法的机密性

分析了压缩感知（CS）的安全性问题,讨论了在攻击者不知道测量矩阵情况下是否 可以有效对信号进行重构的问题,论证了压缩感知可以达到保密性但达不到完善的保密性 。最后联合信道容量和速率失真函数,讨论了压缩感知恢复信号所需测量数据量的下限,并 分析了测量噪声对信号重构性能的影响。     

基于过完备字典稀疏表示的多通道脑电信号压缩感知联合重构

该文基于多通道脑电信号时空特性构建非正交变换过完备字典,准确稀疏表示蕴含时空相关性信息的多通道脑电信号,提高基于时空稀疏贝叶斯学习模型的多通道脑电信号压缩感知联合重构算法性能。实验选用eegmmidb脑电数据库的多通道脑电信号验证所提算法有效性。结果表明,基于过完备字典稀疏表示的多通道脑电信号,能够为多通道脑电信号压缩感知重构算法提供更多的时空相关性信息,比传统多通道脑电信号压缩感知重构算法所得的信噪比值提高近12 dB,重构时间减少0.75 s,显著提高多通道脑电信号联合重构性能。     

基于压缩感知的间歇采样转发干扰方法

针对现代雷达信号带宽大、频率高,相干干扰实现困难的问题,提出了基于压缩感知的间歇采样转发干扰方法.研究了压缩感知基本原理及其对模拟信号处理的实现方法,分析了压缩感知对线性调频信号的间歇采样实现方法及其转发干扰效果.仿真结果表明,压缩感知理论对宽带雷达信号间歇采样具有较好的适用性,对宽带雷达能够实现较好的干扰效果.     

基于压缩感知理论的波段重构方法

针对高光谱图像数据量大、信息冗余多、传输难度大等问题,从波段压缩采样入手,通过采样数据重构出原始波段,提出一种基于压缩感知理论的波段重构方法。压缩感知理论是一种在不遵循奈奎斯特采样定理的情况下,能够高精度重构出原始信号的新型压缩采样理论。由于高光谱图像谱间相关性高,具有很强的稀疏性,故可将压缩感知理论用于高光谱数据的波段重构,仅选择少量波段,便能够重构得到原始高光谱数据。实验结果表明,压缩感知理论能够对高光谱图像波段维进行压缩与重构,并可达到较高的重构比例,同时获得较高的重构效率,且重构数据光谱曲线与原始数据光谱曲线的波形一致度高。     

压缩感知伪随机动态功率信号的电能测量方法

针对压缩感知检测方法不能准确测量伪随机动态测试信号电能量值的问题,该文首先分析了动态测试信号的频域稀疏性,证明动态测试信号满足压缩感知检测的条件;然后采用系统稳态优化的方法,构造了一种确定型压缩感知测量矩阵,证明其符合限制等距特性(RIP)条件,最后提出了一种新型压缩感知动态测试信号电能量的测量方法。实验结果表明：该文压缩感知测量方法的理论相对误差优于传统的采样功率电能测量方法,能够实现m序列动态测试信号的电能量值准确测量。     

基于扩散机制的分布式宽带压缩频谱感知算法

针对认知无线电网络共享频谱资源的特征,本文提出一种基于扩散机制的分布式宽带压缩频谱感知方法。该算法包含两个工作阶段。在第一个阶段,每个认知用户对观测信号进行压缩感知和独立重构,产生本地频谱估计;在第二阶段,各个认知用户根据扩散机制协作更新频谱估计信息,实现最优估计。仿真结果表明,该算法与一致性分布式压缩频谱感知方法相比,可以快速增强认知无线电网络频谱感知能力,可应用于动态拓扑结构的认知无线电网络。     

基于压缩感知的偏振光成像技术研究

偏振成像技术是一种基于目标自身辐射或反射信号中所包含的偏振信息获取物体图像的方法,尤其在人工目标的探测和表面识别方面,相对光强度探测方式具有独特的优势。针对传统的偏振成像技术在复杂的成像环境中成像距离短和成像质量差的缺点,提出了一种基于压缩感知的新型偏振光成像技术。阐述了压缩感知理论的基本原理,构造了合适的采样矩阵和重构算法,设计了具体的成像系统,并通过压缩感知偏振成像实验证明了该成像技术的可行性。空气中实验结果表明,该成像系统能够重构出预先放置目标靶的偏振图像。此外在现有的实验条件基础上讨论并提出了几种改进系统成像质量的措施。     

基于随机采样的超高分辨率成像中快速压缩感知分析

随着近年来超分辨成像技术的发展,基于单分子拟合的超分辨成像方法能够实现纳米尺度的空间分辨率,但这种方法的耗时较长,时间分辨率较差。成像重构时间较长主要受制于成像过程中每帧图像较低的荧光分子密度,所以需要足够多的采样帧数来重构一张图像。文中提出一种利用随机采样的快速压缩感知算法,结合分块压缩感知重构算法,最终能够在高分子密度的条件下获得较快的重构速度及较高的定位精度。     

Adaptive sampling for compressed sensing based image compression

The compressed sensing (CS) theory has been successfully applied to image compression in the past few years as most image signals are sparse in a certain domain. In this paper, we focus on how to improve the sampling efficiency for CS-based image compression by using our proposed adaptive sampling mechanism on the block-based CS (BCS), especially the reweighted one. To achieve this goal, two solutions are developed at the sampling side and reconstruction side, respectively. The proposed sampling mechanism allocates the CS-measurements to image blocks according to the statistical information of each block so as to sample the image more efficiently. A generic allocation algorithm is developed to help assign CS-measurements and several allocation factors derived in the transform domain are used to control the overall allocation in both solutions. Experimental results demonstrate that our adaptive sampling scheme offers a very significant quality improvement as compared with traditional non-adaptive ones.     

基于行列式随机循环的压缩感知测量矩阵研究

压缩感知理论,从信号的自身特性出发,通过变换作用域和线性投影实现对信号的采样和压缩。测量矩阵是该理论中获得最优测量,实现精确重构的关键。本文在介绍常用测量矩阵的基础上,重点研究了结构化测量矩阵。鉴于测量矩阵设计的最重要的原则是降低矩阵元素间的相干性,本文借鉴循环矩阵和广义轮换矩阵的优点,提出了采用均匀随机数对结构化测量矩阵进行随机循环的构造方法。仿真实验表明新矩阵在信号重建上具有更好的性能。     

Novel method of moving target detection for dual-channel was radar based on compressed sensing

We propose a ground moving target detection method for dual-channel Wide Area Sur- veillance （WAS） radar based on Compressed Sensing （CS）. Firstly, the method of moving target detection of the WAS radar is studied. In order to reduce the sample data quantity of the radar, the echo data is randomly sampled in the azimuth direction, then, the matched filter is used to perform the range direction focus. We can use the compressive sensing theory to recover the signal in the Doppler domain. At last, the phase difference between the two channels is compensated to suppress the clutter. The result of the simulated data verifies the effectiveness of the proposed method.     

基于压缩感知的医学图像采样新方法研究

压缩感知基于信号的稀疏性或可压缩性,能够以低于奈奎斯特采样率的频率进行采样,从而直接获得压缩后的采样信号。本文将压缩感知的思想应用于医学图像,对已获得的图像进行压缩采样,以降低医学图像的存储空间。基于正交匹配追踪思想,提出分块双阈值正交匹配追踪方法,根据图像不同区域信息量的不同,采取分块处理并加入采样阈值,针对不同子图像块采取不同采样率,提高了采样效率;同时,在重构时加入判断阈值,可降低重构效果对采样阈值的依赖,减小了由于图像的特殊性或人为经验不足而使得采样阈值选取不当对图像重建质量的影响。实验表明,本文方法能够以较低的采样率实现较高的重构精度。     

压缩感知在多源图像融合中的应用

针对全采样传统图像融合方法中计算量大、时间复杂度高的问题,提出了一种基于压缩感知(CS)理论的多源图像融合模型。为满足一定的稀疏性,将源图像在过完备二维离散余弦变换(DCT)字典上进行稀疏表示,并通过随机观测得到待融合的观测值;在每一图像块上采用基于标准差的方法自适应地计算融合权值,加权合成融合后的观测值,然后利用改进步长的梯度追踪算法求解稀疏系数,得到最终融合图像。实验结果表明：与传统方法相比,提出的融合模型在减少计算量和存储容量的同时,能更好地从源图像中提取信息,获得效果较好的融合图像。     

基于压缩感知的伪随机等效采样信号重构

伪随机等效采样利用采样周期数与采样点数间的互质关系使各采样点均匀复现于同一周期,从而达到较高的等效采样速率。然而为了精确重构出原始信号,需大量采样数据,因此导致采样时间过长,实时性能差。针对上述问题,提出了一种基于压缩感知理论的伪随机等效采样信号重构方法,通过构造伪随机等效采样观测矩阵并选择离散傅里叶变换基建立稀疏重构模型,然后利用压缩感知中的正交匹配追踪算法求解该模型,从而重构出原始信号。仿真实验表明,所提方法在采样点个数40时,重构成功率达99.73%。     

A class of deterministic construction of binary compressed sensing matrices

Compressed Sensing (CS) is an emerging technology in the field of signal processing, which can recover a sparse signal by taking very few samples and solving a linear programming problem. In this paper, we study the application of Low-Density Parity-Check (LDPC) Codes in CS. Firstly, we find a sufficient condition for a binary matrix to satisfy the Restricted Isometric Property (RIP). Then, by employing the LDPC codes based on Berlekamp-Justesen (B-J) codes, we construct two classes of binary structured matrices and show that these matrices satisfy RIP. Thus, the proposed matrices could be used as sensing matrices for CS. Finally, simulation results show that the performance of the proposed matrices can be comparable with the widely used random sensing matrices.     

The high resolution MIMO radar system based on minimizing the statistical coherence of compressed sensing matrix

Compressed Sensing (CS) theory is a great breakthrough of the traditional Nyquist sampling theory. It can accomplish compressive sampling and signal recovery based on the sparsity of interested signal, the randomness of measurement matrix and nonlinear optimization method of signal recovery. Firstly, the CS principle is reviewed. Then the ambiguity function of Multiple-Input Multiple- Output (MIMO) radar is deduced. After that, combined with CS theory, the ambiguity function of MIMO radar is analyzed and simulated in detail. At last, the resolutions of coherent and non-coherent MIMO radars on the CS theory are discussed. Simulation results show that the coherent MIMO radar has better resolution performance than the non-coherent. But the coherent ambiguity function has higher side lobes, which caused a deterioration in radar target detection performances. The stochastic embattling method of sparse array based on minimizing the statistical coherence of sensing matrix is proposed. And simulation results show that it could effectively suppress side lobes of the ambiguity function and improve the capability of weak target detection.     

Performance analysis of support recovery in compressed sensing

Compressed sensing is a new framework to capture sparse signals at sub-Nyquist rate. To guarantee reliable recovery from compressed measurements, the exact reconstruction of the support is necessary. In this letter, we study the probability of exact support reconstruction for maximum likelihood decoding algorithm. The recovery problem is first casted as the one of finding K one-dimensional subspaces containing the maximum energy of the received signal, where K is the sparsity level. Then, the asymptotic probability is developed based on the derived probability density function of the normalized distances between the received signal and all possible subspaces.     

一种低速采样的协同宽带频谱感知方法

频谱感知是通信系统抗干扰和智能化的关键能力.针对认知无线电系统窄带频谱感知技术受制于数模转换器件的发展水平,难以解决认知无线电系统宽带、实时频谱感知的问题,提出一种多节点协作的认知无线电系统宽带频谱感知方法.该方法设计由多个认知节点对目标频段执行次奈奎斯特采样来降低采样速率,采用能量检测方式对采样矢量进行集中式融合判决,实现宽频段范围内干扰信号的谱定位和判断,降低各个感知节点的采样速率,支撑认知网络系统构建高实时、宽频带频谱感知的能力.计算机仿真试验结果表明,所提方法达到90％检测概率时压缩比要求为0.025,具有可靠性与有效性.