基于并行BP神经网络的近红外光断层图像重建方法基础研究

提出一种基于并行BP神经网络的近红外光断层成像(Near-infrared optical tomography,NIR OT)图像重建算法,利用BP神经网络来表征生物组织内部光学参数的空间分布和边界光强之间的非线性映射关系.该方法将一个复杂的模型分解成简单的模型分别建立并行的神经网络.利用Femlab软件完成基于有限元的稳态扩散方程的两个简单模型的正向问题求解,根据提出的平均优化散射系数和正向问题训练的大量数据集合,建立并训练并行神经网络,通过对两个网络结果的分析,实现快速获得更复杂模型的光学参数的重构.算法能够快速识别特异组织的位置和准确反映热疗过程中生物组织的优化散射系数的变化趋势.     

A Novel Finite-Element-Based Algorithm for Fluorescence Molecular Tomography of Heterogeneous Media

The knowledge of optical properties distribution of heterogeneous media has significant impact on the reconstructed fluorescence image quality in fluorescence molecular tomography (FMT). In this study, a novel finite-element-based algorithm for FMT of heterogeneous media is proposed. In the algorithm, optical properties are reconstructed using the conjugate gradient method. A modified method based on reverse differential scheme is deduced for calculating the gradient when the detector points are not restricted on boundary nodes. With the recovered optical properties, linear relationship between known surface measurements of emission light and unknown fluorescence yield is then obtained. FMT reconstruction is implemented by combining algebraic reconstruction technique (ART) and Landweber iteration method. With initial value provided by ART, Landweber iteration method improves the quantification smoothly with small step length between neighboring iterations. The algorithm was evaluated using phantoms of different heterogeneity configurations. Results show that the reconstructed fluorescence yield is insensitive to various degrees of heterogeneity for the proposed algorithm. In contrast, when assuming homogeneous optical properties, it shows that more underestimation of optical properties results in more underestimation of the reconstructed fluorescence yield. Fast computation speed of the proposed algorithm is also demonstrated in this study.     

Diffuse optical tomography of highly heterogeneous media

We investigate the performance of diffuse optical tomography to image highly heterogeneous media, such as breast tissue, as a function of background heterogeneity. To model the background heterogeneity, we have employed the functional information derived from Gadolinium-enhanced magnetic resonance images of the breast. We demonstrate that overall image quality and quantification accuracy worsens as the background heterogeneity increases. Furthermore we confirm the appearance of characteristic artifacts at the boundaries that scale with background heterogeneity. These artifacts are very similar to the ones seen in clinical examinations and can be misinterpreted as actual objects if not accounted for. To eliminate the artifacts and improve the overall image reconstruction, we apply a data-correction algorithm that yields superior reconstruction results and is virtually independent of the degree of the background heterogeneity.     

基于耦合梯度神经网络的光学层析图像重建

为克服光学层析图像重建的病态性,通常在重建过程中加入先验信息.本文采用含有二值线过程的Gibbs分布作为图像的先验模型,该模型具有保留清晰边缘的全局平滑特性.由于重建目标函数是连续变量和二值离散变量的混合体,常规的优化算法无法实现.为此,提出了一种基于耦合梯度神经网络的优化方法.优化过程中,能量函数关于光学参数的梯度计算是关键,本文提出一种基于梯度树的梯度求解方法.对吸收系数和散射系数的重建结果表明:该方法可高效地重建光学层析图像;线过程的引入可以改善重建的病态特性,提高图像的重建质量.     

基于水体退化函数补偿方法的水下傅里叶单像素成像

水下光学成像的探测环境相对复杂。前向散射、后向散射和吸收极大地降低了水下光学成像的成像质量。单像素成像因其较高的抗噪声性而被认为是一种适合于水下光学成像的技术。对于水下单像素成像系统,目前存在的问题是需要采用结构光进行照明,当散斑在水下传播时,前向散射会使预先生成的散斑发生畸变。因此,重建结果的分辨率降低,重建结果模糊不清。为了减小前向散射对单像素成像系统的影响,对傅里叶单像素成像的重建过程进行改进。在空间谱域傅里叶单像素成像系统的水下退化函数进行估计,然后根据估计的退化函数实现目标空间谱反演。最后利用傅里叶变换对反演后的目标空间谱进行变换,最终获得目标的强度图像。理论分析和实验结果证明了该方法的有效性。利用该方法,能够有效地减小前向散射对成像质量的影响,提高了水下傅里叶单像元成像的重建结果质量。     

基于自适应鲁棒性光流的差错掩盖

编码数字视频序列经过受噪声影响的信道传输时,通常会出现图像信息丢失。该文提出一种基于自适应鲁棒性光流的差错掩盖方法,作为解码端的工具解决这样的问题。该文利用光流技术能有效获取物体运动估计的特性,对丢失块进行逐像素点的恢复,既避免了图像模糊,又消除了块效应。在光流的估计中,自适应地调整了目标泛函中的数据保持项与空间连贯项之间的关系,并引入Lorentz函数来构造目标泛函,提高了光流的鲁棒性。仿真结果表明,该文提出的方法无论在主观视觉评价,还是在客观的数值标准下,都能比现有的误差掩盖方法恢复出质量更好的图像。     

Image reconstruction with locally adaptive sparsity and nonlocal robust regularization

Sparse representation based modeling has been successfully used in many image-related inverse problems such as deblurring, super-resolution and compressive sensing. The heart of sparse representations lies on how to find a space (spanned by a dictionary of atoms) where the local image patch exhibits high sparsity and how to determine the image local sparsity. To identify the locally varying sparsity, it is necessary to locally adapt the dictionary learning process and the sparsity-regularization parameters. However, spatial adaptation alone runs into the risk of over-fitting the data because variation and invariance are two sides of the same coin. In this work, we propose two sets of complementary ideas for regularizing image reconstruction process: (1) the sparsity regularization parameters are locally estimated for each coefficient and updated along with adaptive learning of PCA-based dictionaries; (2) a nonlocal self-similarity constraint is introduced into the overall cost functional to improve the robustness of the model. An efficient alternative minimization algorithm is present to solve the proposed objective function and then an effective image reconstruction algorithm is presented. The experimental results on image deblurring, super-resolution and compressive sensing demonstrate that the proposed image reconstruct method outperforms many existing image reconstruction methods in both PSNR and visual quality assessment.     

基于星图模拟的星空目标提取

提出了一种将星图模拟技术应用于星空背景空间目标提取的方法,建立了地基光测设备的星图模拟模型,推导了光轴指向计算和星体到焦平面成像的坐标转换。利用模拟星图生成掩模,对实拍星图处理去除恒星背景。实验证明利用模拟星图能够有效去除恒星背景提取空间目标。     

Acceleration of optoacoustic model-based reconstruction using angular image discretization

Optoacoustic tomography has recently demonstrated powerful performance in small animal imaging and initial clinical trials in terms of the high spatial resolution, versatile contrast, and dynamic imaging capabilities it can provide. Yet, the current optoacoustic image reconstruction methods are usually based on inaccurate forward modelling approaches or otherwise demand a high computational cost, which imposes certain practical limitations and hinders image quantification. Herein, we introduce a new method for accelerating optoacoustic reconstructions, based on angular image discretization of the forward model solution. The method is particularly suitable for accurate image reconstruction with arbitrary meshes and space-dependent resolution, while it can also readily account for small speed of sound variations without compromising the calculation speed. It is further anticipated that the new approach will greatly facilitate development of high performance 3-D optoacoustic reconstruction methods.     

Two-dimensional hybrid element image reconstruction for TMillumination

A Newton's iterative scheme for electromagnetic imaging is proposed for reconstruction of the dielectric properties of inhomogeneous, lossy bodies with arbitrary shape. The algorithm is based on a finite element (FE) representation which is coupled to a boundary element (BE) formulation for the forward solution of the electric fields; together these are termed the hybrid element method. It utilizes FE discretization of only the area of interest while incorporating the RE method to match the conditions of the homogeneous background region extending to infinity. This paper presents image reconstruction for the 2D TM polarization case where two classes of dielectric distributions are studied which demonstrate the flexibility of this method along with some of the difficulties associated with larger imaging problems     

Estimation of images and nonrigid deformations in gated emission CT

In this paper, we propose and test a new iterative algorithm to simultaneously estimate the nonrigid motion vector fields and the emission images for a complete cardiac cycle in gated cardiac emission tomography. We model the myocardium as an elastic material whose motion does not generate large amounts of strain. As a result, our method is based on minimizing an objective function consisting of the negative logarithm of a maximum likelihood image reconstruction term, the standard biomechanical model of strain energy, and an image matching term that ensures a measure of agreement of intensities between frames. Simulations are obtained using data for the four-dimensional (4-D) NCAT phantom. The data models realistic noise levels in a typical gated myocardial perfusion SPECT study. We show that our simultaneous algorithm produces images with improved spatial resolution characteristics and noise properties compared with those obtained from postsmoothed 4-D maximum likelihood methods. The simulations also demonstrate improved motion estimates over motion estimation using independently reconstructed images.     

基于全变差的加权最小二乘法PET图像重建

传统的加权最小二乘法、惩罚项加权最小二乘法虽然能够重建得到较好质量的图像,但在欠采样的条件下不能很好的拟制噪声.全变差作为正则项已广泛用于图像重建中,利用图像稀疏的先验知识能够在欠采样的条件下很好的重建图像.本文结合加权最小二乘法和全变差的优点,构造了基于全变差正则项的加权最小二乘法目标函数,运用交替求解的方法,将目标函数分解为求解二次优化和全变差正则化的优化问题,并分别用超松弛迭代方法和梯度下降法求解这两个优化问题.采用Zubal模型对该算法与传统算法进行仿真验证比较,并用相关系数、方差、信噪比等参数描述图像重建质量.结果表明在欠采样条件下,该算法能够更好的拟制噪声,重构效果比传统的有明显地提高.     

一种水下图像增强算法的实现和仿真

〗由于水下环境的特殊性和复杂性,使得水下图像的质量差、图像对比度低。文中给出了一种基于空间域的图像增强方法,该算法利用均值算法估计水下图像背景,从原水下图像衰减背景图像,再对衰减背景之后的图像进行改进的图像锐化处理。通过对算法仿真结果的分析可知,处理后的图像整体对比度明显提升,同时使得目标的边缘更加清晰。     

一种分析光纤共聚焦系统轴向光强的方法

在携带光纤或传像束的共聚焦系统中,光场在光纤或传像束的端面要发生改变,不能用通常的空间线性不变方法分析系统成像过程。根据单模光纤的传输特性,获得光纤的有效点扩散函数,提出利用光纤空间平均函数描述系统空间变成像过程的方法,模拟出光纤共聚焦系统的轴向光强分布,对比由光波动理论所得结果,显示该方法在光纤输出光斑的有效尺寸较小时能用于分析携带光纤或传像束的共聚焦系统且更简便。     

基于改进稀疏表示正则化的SR重建算法

基于视觉的无人机自主导航过程中,对航路点进行准确识别是引导无人机朝着航路点方向精确飞行的关键。然而,当无人机到达航路点识别距离后,由于机载图像传感器受天气因素及成像过程中的脱焦、衍射等现象影响,常导致获取到的航拍图像模糊、空间分辨率较低,从而直接影响了后续航路点识别的精度。针对这一问题,提出了一种改进稀疏表示正则化的航拍图像超分辨率重建算法。首先,基于稀疏表示正则化框架,利用自回归和非局部相似约束构建目标函数的正则化项;其次,根据图像局部方差能有效区分图像的边缘区域和平滑区域这一特性,自适应地选取正则化参数得到超分辨率重建模型中的目标函数;最后,使用MM (Majorization-Minorization) 算法求解目标函数的凸优化问题,得到重建后的高分辨率图像。实验结果表明:与传统的正则化SR重建算法相比,文中算法能够有效的提高航拍图像的空间分辨率,使得重建后的图像包含了更多的特征细节信息,这为航路点识别提供了帮助。     

视觉检测系统LED光源照度优化方法

针对当前视觉检测系统LED光源照度优化研究中存在的照度效果评价因素单一、照度优化方法通用性不足等问题,以芯片封装质量视觉检测为例,提出一种基于改进樽海鞘算法的LED光源照度优化方法。该方法在单个LED光源照度数学模型基础上,建立标准条形LED阵列光源照度数学模型,获取条形LED阵列在任意空间位姿与被测面的照度值;基于照度均匀度、照度梯度变化与对中度、平均照度、目标与背景区分度等因素建立平面照度效果评价函数;提出改进樽海鞘算法,通过改进算法收敛系数、速度、领导者与追随者位置等更新策略,增强区域搜索的多样性;应用改进樽海鞘算法对平面照度效果评价函数进行优化求解,获取具有最优照度效果的空间位姿参数。实验结果表明:考察优化区域的相对照度分布,文中提出的LED光源照度优化方法所得照度分布与实际测量所得照度分布结果基本一致,目标区域理论照度均匀度在98.78%以上,误差在5.57%以内。因此文中提出方法优化目标合理,可用于视觉检测系统具有最优照度效果时光源位姿信息参数的获取。     

Adaptive reconstruction technique for the lost information of the rectangular image area

The adaptive reconstruction for the lost information of the rectangular image area is very important for the robust transmission and restoration of the image. In this paper, a new reconstruction method based on the Discrete Cosine Transform (DCT) domain has been put forward. According to the low pass character of the human visual system and the energy distribution of the DCT coefficients on the rectangular boundary, the DCT coefficients of the rectangular image area are adaptively selected and recovered. After the Inverse Discrete Cosine Transform (IDCT), the lost information of the rectangular image area can be reconstructed. The experiments have demonstrated that the subjective and objective qualities of the reconstructed images are enhanced greatly than before.     

深空可见光图像中弱小运动目标实时检测

 针对深空可见光图像背景特征与目标特性,提出了迭代质心的方法自适应地搜索恒星灰度质心作为特征点并构造基于恒星空间分布的特征模型,实现亚像素级精度的图像序列配准,通过8-邻域联通聚类分析的路径判别法,解决了弱小运动目标实时检测.     

采用高频能量的CT几何参数自标定方法

针对锥束CT系统中几何参数失配引起几何伪影的问题,提出了一种采用空域高频能量的几何伪影自校正算法。该算法以重建图像的空域高频能量为目标函数建立优化模型,通过单纯形法迭代求解使空域高频能量最大的几何参数最优解。利用投影图像的特性提取部分参数作为输入初值,减小算法搜索范围。并采用图形处理器（Graphic Processing Unit,GPU）对自校正过程中的图像重建并行加速,减少重建时间,提高校正速度。实验结果表明:该算法具有较高的求解精度,最大相对误差不超过5%,对重建图像中的几何伪影有较好的校正效果。同时,在不影响精度的情况下减少了迭代次数,算法执行效率提高了18.8%。     

Application of the multilevel single-linkage method toone-dimensional electromagnetic inverse scattering problem

An inverse scattering method for the reconstruction of the permittivity and conductivity profiles of a multilayered medium and for that of the impedance profile of a nonuniform transmission line is proposed. The inversion is based on the global minimization of an objective function by the multilevel single-linkage method. The objective function is defined as the mean-square error between the measured data and the data obtained from the solution of the forward problem. An exact formulation for the gradient of the objective function in closed form is derived. The necessary condition for the unique solution of the inverse problem of a nonuniform transmission line is discussed. Reconstruction examples using both experimental and noisy synthetic data are presented