基于全变分正则化的高能闪光照相消模糊图像重建算法

针对高能闪光照相投影图像消模糊难度大的问题,提出了一种基于全变分正则化的消模糊图像重建算法,该算法根据闪光照相的成像特点,将客体的纵向截面作为一个整体来进行建模,并在重建方程中考虑了模糊因素,然后采用全变分范数作为正则项,构建了用于消模糊图像重建的展平泛函,将消模糊图像重建问题转化为能量泛函极小化问题,通过固定点迭代算法求解图像重建问题的最小化解。数值模拟结果表明：该算法由于考虑了闪光照相成像时的图像模糊因素,在重建时能够较好地消除模糊对重建结果的影响,在抑制噪声的同时能较好地保持图像的边缘信息,有利于提高重建图像的质量。     

基于全变分正则化的高能闪光照相消模糊图像重建算法

 针对高能闪光照相投影图像消模糊难度大的问题,提出了一种基于全变分正则化的消模糊图像重建算法,该算法根据闪光照相的成像特点,将客体的纵向截面作为一个整体来进行建模,并在重建方程中考虑了模糊因素,然后采用全变分范数作为正则项,构建了用于消模糊图像重建的展平泛函,将消模糊图像重建问题转化为能量泛函极小化问题,通过固定点迭代算法求解图像重建问题的最小化解。数值模拟结果表明：该算法由于考虑了闪光照相成像时的图像模糊因素,在重建时能够较好地消除模糊对重建结果的影响,在抑制噪声的同时能较好地保持图像的边缘信息,有利于提高重建图像的质量。     

基于双阈值Huber范数估计的图像正则化超分辨率算法

多帧图像超分辨率算法利用图像间的互补信息, 可以从一系列具有亚像素位移的低分辨率影像数据中重建出高分辨率图像. 在众多超分辨率算法中, 正则化方法以其求解病态问题的有效性而被广泛应用, 但在此类方法中, 最优估计算子的估计准确度对最后的重建结果有着较大的影响. 本文在现有正则化超分辨率重建算法的基础上, 提出了一种基于双阈值Huber范数的极大似然估计算子, 可以提高Huber范数对于阈值取值的容忍性和算子估计精度; 并给出了基于该算子的正则化超分辨率算法的迭代公式. 通过对仿真图像进行重建, 结果表明算法可有效地抑制各种噪声并保证重建效果; 同时将此算法应用于实际图像的超分辨率重建, 有效地提高了目标影像的空间分辨率. 关键词： 图像处理 图像重建 超分辨率 正则化     

基于自适应正则化的超分辨率重建方法

传统正则化超分辨重建得到的图像往往存在过度平滑或伪信息残留的问题,结合超分辨重建模型对重建图像伪信息的产生进行了分析,针对传统方法的不足提出了基于图像区域信息自适应的正则化方法,通过图像的区域信息将图像划分为平滑区与非平滑区域,对不同区域选用不同的先验模型进行约束。同时考虑人眼的视觉感知特性,结合区域信息实现正则化参数的自适应选取。实验结果表明该方法在抑制重建图像伪信息的同时能有效保护细节,效果要优于传统方法与单一的先验模型约束,对于红外与可见光图像重建效果的提升提供了一定的理论参考。     

基于Gadgetron平台的多GPU分布式磁共振图像重建

为了满足磁共振成像（MRI）临床扫描的需求,磁共振图像重建算法的开发一直在不断进行.目前广泛使用的算法实现方式是利用中央处理器（CPU）对磁共振扫描数据进行数学变换得到图像,随着算法复杂度的提升,计算性能问题逐渐显露.利用CPU在大数据量下执行复杂算法时,计算并行性的缺失以及运算中产生的海量数据的存储负荷会导致计算变得极为缓慢,使得一些算法因为重建时间过长,在临床上面临难以推广的问题,也制约了基础研究中新算法的研发.本文设计并实现了一种新的重建算法执行方式,利用Gadgetron磁共振软件重建平台在多核CPU基础上搭载多块图形处理器（GPU）,将磁共振图像重建以分布式并行计算方式实现,并以重建耗时较长的3D径向数据采集Stack of Star（SOS）的图像重建为实例,展示这种重建的实现方法能以相对低廉的硬件成本极大提升重建的速度.     

并行MRI图像重建算法比较及软件实现

首先介绍了不加速的并行MRI图像重建方法,然后对加速的并行MRI的4种图像重建算法进行了比较,得出结论：加速因子相同时,重建质量上,GRAPPA和SENSE的重建质量最好,SMASH的重建质量次之, PILS算法对线圈位置要求极高,重建质量最差;重建速度上,SMASH的重建速度最快,其次是SENSE和PILS,GRAPPA的重建速度最慢. 当加速因子变大时,所有算法重建质量都变差. 最后介绍了算法实现软件,该软件可以读入原始数据,显示数据采集轨迹,计算线圈灵敏度,选择图像重建方法,分析和比较重建图像质量. 该软件为我国在MRI成像领域提供了一个学习和进一步研究图像重建算法的有力工具.     

基于支撑先验与深度图像先验的无预训练磁共振图像重建方法

基于深度学习的磁共振成像(magnetic resonance imaging, MRI)方法需要大规模、高质量的病患数据样本集进行预训练.然而,由于病患隐私及设备等因素限制,获取大规模、高质量的磁共振数据集在实际临床应用中面临挑战.本文提出一种新的基于深度学习的欠采样磁共振图像重建方法,该方法无需预训练、不依赖训练数据集,而是充分利用待重建的目标MR图像的结构先验和支撑先验,并将其引入深度图像先验(deep image prior, DIP)框架,从而削减对训练数据集的依赖,提升学习效率.基于参考图像与目标图像的相似性,采用高分辨率参考图像作为深度网络输入,将结构先验信息引入网络;将参考图像在小波域中幅值大的系数索引集作为目标图像的已知支撑集,构造正则化约束项,将网络训练转化为网络参数的最优化求解过程.实验结果表明,本文方法可由欠采样k空间数据重建得到更精确的磁共振图像,且在保留组织特征、细节纹理方面具有明显优势.     

模糊条码图像的正则化复原算法

由条码扫描仪获得条码图像的过程可以用理想条码信号与扫描仪光学系统点扩散函数的卷积模型来描述。反卷积是消除由光学系统点扩散带来的模糊现象的最好办法。为克服反卷积的病态问题,研究了反卷积的正则化方法;针对条码信号的特点,构建了适合于条码信号复原的惩罚项,提出了条码信号的正则化复原算法及其适合于计算机运算的迭代算法。通过实验研究了算法在不同情况下的抗干扰能力。实验结果表明,正则化条码信号复原算法在消除系统点扩散函数的影响的同时能够很好地抑制噪声。     

一种基于约束共轭梯度的闪光照相图像重建算法

针对闪光照相系统成像信噪比低的特点,提出了正则化预优约束共轭梯度算法（RPCCG）。RPCCG算法在闪光照相重建方程中引入Tikhonov正则化准则,利用预优约束共轭梯度法迭代求图像重建的最优解。数值试验表明,采用最小二乘+平滑准则的RPCCG算法是一种具有较高的抗噪能力的有效闪光照相图像重建算法,具有良好的收敛性和稳定性以及较高的重建精度。     

一种基于约束共轭梯度的闪光照相图像重建算法

 针对闪光照相系统成像信噪比低的特点,提出了正则化预优约束共轭梯度算法（RPCCG）。RPCCG算法在闪光照相重建方程中引入Tikhonov正则化准则,利用预优约束共轭梯度法迭代求图像重建的最优解。数值试验表明,采用最小二乘+平滑准则的RPCCG算法是一种具有较高的抗噪能力的有效闪光照相图像重建算法,具有良好的收敛性和稳定性以及较高的重建精度。     

无需线搜索的并行非线性共轭梯度法

给出了一类无需线搜索的无约束最优化并行算法--并行非线性共轭梯度法(NLS-PNCG),用一个固定的公式来计算搜索步长,较常用的共轭梯度法计算量小.并且证明了目的的全局收敛性,给出了数值试验,结果显示NLS-PNCG优于线搜索的非线性共轭梯度法(NCG).     

Conjugate Gradient method for finding fundamental solitary waves

The Conjugate Gradient method (CGM) is known to be the fastest generic iterative method for solving linear systems with symmetric sign definite matrices. In this paper, we modify this method so that it can find fundamental solitary waves of nonlinear Hamiltonian equations. The main obstacle that such a modified CGM overcomes is that the operator of the equation linearized about a solitary wave is not sign definite. Instead, it has a finite number of eigenvalues on the opposite side of zero than the rest of its spectrum. We present versions of the modified CGM that can find solitary waves with prescribed values of either the propagation constant or power. We also extend these methods to handle multi-component nonlinear wave equations. Convergence conditions of the proposed methods are given, and their practical implications are discussed. We demonstrate that our modified CGMs converge much faster than, say, Petviashvili’s or similar methods, especially when the latter converge slowly.     

Non-Iterative Regularized reconstruction Algorithm for Non-CartesiAn MRI: NIRVANA

We introduce a novel noniterative algorithm for the fast and accurate reconstruction of nonuniformly sampled MRI data. The proposed scheme derives the reconstructed image as the nonuniform inverse Fourier transform of a compensated dataset. We derive each sample in the compensated dataset as a weighted linear combination of a few measured k-space samples. The specific k-space samples and the weights involved in the linear combination are derived such that the reconstruction error is minimized. The computational complexity of the proposed scheme is comparable to that of gridding. At the same time, it provides significantly improved accuracy and is considerably more robust to noise and undersampling. The advantages of the proposed scheme makes it ideally suited for the fast reconstruction of large multidimensional datasets, which routinely arise in applications such as f-MRI and MR spectroscopy. The comparisons with state-of-the-art algorithms on numerical phantoms and MRI data clearly demonstrate the performance improvement.     

基于概率加权共轭梯度算法的混凝土超声波层析成像

在混凝土层析成像中,为了提高反演的准确性和计算效率,针对共轭梯度算法提出一种加权算法——概率加权共轭梯度算法.新算法不同于常规的加权算法,权重是加在成像单元上而不是方程上.为取得较好的权重因子和较好的迭代初始值,采用IART算法的权重和迭代初始值的选取方法.模拟算例和混凝土试验均表明这种加权算法的可行性.     

Coherent Beam Combining of Three Watt-Level Fiber Amplifiers Using a DSP-Based Stochastic Parallel Gradient Descent Algorithm

We present an experimental study on coherent beam combining of three watt-level fiber amplifiers using a stochas- tic parallel gradient descent （SPGD） algorithm. Phase controlling is performed by running the SPGD algorithm on a digital-signal-processor （DSP） chip with a voltage updating rate of 16500 times per second. Energy encircled in the target pinhole is 2.62 times more than that in an open loop. The combining efficiency is as high as 87%.     

高能X光照相中密度重建的共轭梯度算法

通过对高能X光照相过程的物理分析,得到了辐射照相的点扩展函数,给出了估价函数的表达式.并以典型的法国试验客体模型为例,将共轭梯度算法应用到高能X光照相的密度重建中,得到了较满意的结果.这种方法从模拟辐射照相过程出发,将辐射照相的模拟结果与实际测量结果做比较并找出最大可能解.可在一定程度上克服其它密度重建算法通过反卷积过程消除模糊效应所引起的不确定性.     

Parallel MRI at microtesla fields

Parallel imaging techniques have been widely used in high-field magnetic resonance imaging (MRI). Multiple receiver coils have been shown to improve image quality and allow accelerated image acquisition. Magnetic resonance imaging at ultra-low fields (ULF MRI) is a new imaging approach that uses SQUID (superconducting quantum interference device) sensors to measure the spatially encoded precession of pre-polarized nuclear spin populations at microtesla-range measurement fields. In this work, parallel imaging at microtesla fields is systematically studied for the first time. A seven-channel SQUID system, designed for both ULF MRI and magnetoencephalography (MEG), is used to acquire 3D images of a human hand, as well as 2D images of a large water phantom. The imaging is performed at 46 mu T measurement field with pre-polarization at 40 mT. It is shown how the use of seven channels increases imaging field of view and improves signal-to-noise ratio for the hand images. A simple procedure for approximate correction of concomitant gradient artifacts is described. Noise propagation is analyzed experimentally, and the main source of correlated noise is identified. Accelerated imaging based on one-dimensional undersampling and 1D SENSE (sensitivity encoding) image reconstruction is studied in the case of the 2D phantom. Actual threefold imaging acceleration in comparison to single-average fully encoded Fourier imaging is demonstrated. These results show that parallel imaging methods are efficient in ULF MRI, and that imaging performance of SQUID-based instruments improves substantially as the number of channels is increased.     

Experimental Investigation of Heat Transfer Coefficient from the Impingement of a Slot Jet Using Conjugate Gradient Method with Adjoint Equation

The convective heat transfer coefficients resulting from a slot jet on a plane surface have been estimated by using the real measured temperatures in the plate and inverse method. In this study, the inverse method used the conjugate gradient method with an adjoint equation. Distributions of the local heat transfer coefficient on the impingement surface were determined for various Re and H/Dh. It was found the heat transfer coefficients generally tended to decrease with increasing separation distance and to increase with an increase in Reynolds number. This presented method is able to estimate the variation of the local Nusselt number with time.     

Gradient hysteresis in MRI and NMR experiments

In this paper, we describe a gradient hysteresis effect that can modulate the current in gradient coils during MRI and NMR experiments. A simple pulse sequence is presented for the purpose of evaluating the resulting changes in the accumulated phase. Additionally, the nature of the gradient pulse shape changes is described. These experiments will be of interest to MRI and NMR scientists who are developing pulse sequences requiring precision gradient performance or who are currently seeking the source of unexplained NMR artifacts.     

A Weak Form of the Conjugate Gradient FFT Method for Two-Dimensional Elastodynamics

The problem of two-dimensional scattering of elastic waves by an elastic inclusion can be formulated in terms of a domain integral equation, in which the grad-div operator acts on a vector potential. The vector potential is the spatial convolution of a Green's function with the product of the density and the displacement over the domain of interest. A weak form of the integral equation for the unknown displacement is obtained by testing it with rooftop functions. This method shows excellent numerical performance.