Iterative ultrasonic signal and image deconvolution for estimation of the complex medium response

The ill‐conditioned inverse problem of estimating ultrasonic medium responses by deconvolution of RF signals is investigated. The primary difference between the proposed method and others is that the medium response function is assumed to be complex‐valued rather than restricted to being real‐valued. Derived from the complex medium model, complex Wiener filtering is presented, and a Hilbert transform related limitation to inverse filtering type methods is discussed. We introduce a nonparametric iterative algorithm, the least squares method with point count regularization (LSPC). The algorithm is successfully applied to simulated and experimental data and demonstrates the capability of recovering both the real and imaginary parts of the medium response. The simulation results indicate that the LSPC method can outperform Wiener filters and improve the resolution of the ultrasound system by factors as high as 3.7. Experimental results using a single element transducer and a conventional medical ultrasound system with a linear array transducer show that despite the errors in pulse estimation and the noise in the RF signals, excellent results can be obtained, demonstrating the stability and robustness of the algorithm. © 2006 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 15, 266–277, 2005     

Comparison of numerical methods for identification of viscoelastic line spectra from static test data

Viscoelastic line spectra are identified from creep or relaxation data of static experiments with different numerical methods, which may or may not depend on additional informations, to be provided by the user, about the unknown parameters. If the least square method is applied, a non‐linear optimization problem with non‐negative constraints on the parameters has to be solved. Its solution can be achieved directly by using a gradient‐based optimization algorithm like the projected Newton method of Bertsekas. However, appropriate starting values for the unknown parameters must be chosen. The problem can be alleviated by dividing the identification task into three successive steps, based on the Tschebyscheff approximation and the quadratic optimization method by Wolfe. Alternatively, the identification task can be reduced to a quadratic optimization problem, if the user provides additional informations about the distribution of the respondance times of the spectra. The windowing‐method of Emri and Tschoegl is based on this assumption. If the line spectrum is assumed to have equally distributed spectrum lines on the logarithmic axis, the identification problem can also be solved by standard regularization techniques, like the truncated singular value decomposition or the Tikhonov regularization. The choice of qualified respondance times as additional information requires some experience with the identification task at hand. Its results may be improved after several reruns of the algorithms. Various applications of the methods to test and experimental data are given and a comparison of their performance is discussed. Copyright © 2005 John Wiley & Sons, Ltd.     

Two-point higher-order BVPs with singularities in phase variables

The existence of solutions for singular higher-order differential equations with the Lidstone or the (n,p) boundary conditions is proved. The right-hand sides of differential equations can have singularities in the zero value of their phase variables and so higher derivatives of solutions changing their signs can pass through these singularities. Proofs are based on the method of a priori estimates, the degree theory arguments and on Vitali's convergence theorem.     

Regularization Techniques for Numerical Approximation of PDEs with Singularities

The rate of convergence for numerical methods approximating differential equations are often drastically reduced from lack of regularity in the solution. Typical examples are problems with singular source terms or discontinuous material coefficients. We shall discuss the technique of local regularization for handling these problems. New numerical methods are presented and analyzed and numerical examples are given. Some serious deficiencies in existing regularization methods are also pointed out.     

一种改进的正则图象

把正则图象的误差分成两部分：一部分为逼近误差，它是和原图象相关的确定型误差，另一部分为噪声传递误差，它是和噪声相关的随机型误差．通过正则图象代替原图象，求得逼近误差的近似值，然后将逼近误差近似值补给正则解，得到改进型正则解．对正则解以及逼近误差的计算，提出了快速算法，并有严密的理论推导．实验结果表明改进型正则图象的恢复效果优于传统正则解．在一敷的正则化方法中，可以借鉴本文提到的方法改进正则解．     

一种新颖的径向基函数(RBF)网络学习算法

以提高RBF网络泛化能力为着眼点，提出了一种新型的网络结构自适应学习算法．该算法采用衰减聚类半径的聚类算法来确定初始的隐层结构，然后通过调整包含样本类别信息的扩展聚类不纯度来修正隐层结构，直至满足所有扩展聚类不纯度均小于等于不纯度均值以及所有扩展聚类方差均不超过方差均值这两个条件．这样就确定了隐层的最终结构．在确定隐层结构之后，采用反向传播算法来训练隐层与输出层之间的连接权重．经双螺旋线问题仿真试验验证，该算法确实具有较强的泛化能力．     

一种图象恢复效果接近最优的正则化方法

提出一种新的技术，它自适应地选取正则化参数以取得较理想的恢复效果．利用小波变换，分析正则化算子和正则化参数对图象残差的各子频段能量的影响．在本文条件下，我们论证正则化算子取拉普拉斯算子比取恒等算子恢复性能好,并且预测噪声能量．实验结果表明本文提出的方法不需要知道噪声能量，也能够自适应地确定正则化参数并且恢复性能比传统的方法好，恢复效果非常接近最优恢复．     

Isotropic Energies, Filters and Splines for Vector Field Regularization

The aim of this paper is to propose new regularization and filtering techniques for dense and sparse vector fields, and to focus on their application to non-rigid registration. Indeed, most of the regularization energies used in non-rigid registration operate independently on each coordinate of the transformation. The only common exception is the linear elastic energy, which enables cross-effects between coordinates. Cross-effects are yet essential to give realistic deformations in the uniform parts of the image, where displacements are interpolated.In this paper, we propose to find isotropic quadratic differential forms operating on a vector field, using a known theorem on isotropic tensors, and we give results for differentials of order 1 and 2. The quadratic approximation induced by these energies yields a new class of vectorial filters, applied numerically in the Fourier domain. We also propose a class of separable isotropic filters generalizing Gaussian filtering to vector fields, which enables fast smoothing in the spatial domain. Then we deduce splines in the context of interpolation or approximation of sparse displacements. These splines generalize scalar Laplacian splines, such as thin-plate splines, to vector interpolation. Finally, we propose to solve the problem of approximating a dense and a sparse displacement field at the same time. This last formulation enables us to introduce sparse geometrical constraints in intensity based non-rigid registration algorithms, illustrated here on intersubject brain registration.     

RKRLS及在混炼胶质量建模与预报中的应用研究

对递推最小二乘(RLS)进行了非线性的核(kernel)变换,并采用正则化技术改写了目标范函,提出了一种正则核变换递推最小二乘(regularized kernel,RLS)算法.获得了RKRLS模型的系数和误差表达式,分析了算法的推广能力并证明了KeYnel RLS算法为其特例,进而导出了RKRLS算法在限定、增长和缩减记忆三种不同模式下的递推公式均无需进行求逆计算.RKRLS算法具有三个特性:小样本、可控的推广能力和速度快,因而非常适合于工业应用场合.通过对橡胶混炼质量的门尼指标进行建模和预测分析表明,本算法具有较好的跟踪预测性能.     

电容层析成像图像重建的总变差正则化算法

针对电容层析成像（ECT）逆问题解的不适定性,本文提出一种基于总变差（total variation,TV）正则化的图像重建算法。同传统的2范数Tikhonov正则化方法相比,该算法（基于1范数正则化）不仅保证了逆问题求解的稳定性,而且提高了对介质非连续分布的区域成像的分辨能力,具有良好的保边缘性。仿真及实验结果表明,该算法在重建图像质量和重建速度两方面均具有优势。