3D velocity from 3D Doppler radial velocity

We present local least squares and regularization frameworks for computing 3D velocity (3D optical flow) from 3D radial velocity measured by a Doppler radar. We demonstrate the performance of our algorithms quantitatively on synthetic radial velocity data and qualitatively on real radial velocity data, obtained from the Doppler radar at Kurnell Radar station, Botany Bay, New South Wales, Australia. Radial velocity can be used to predict the future positions of storms in sequences of Doppler radar datasets.© 2005 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 15, 189–198, 2005     

叠前地震数据的规则化

本文通过对动校正前积分法DMO原理的探讨，在借鉴Canning(1996)提出的利用DMO实现三维数据规则化的方法和Biondi(1998)提出的三维叠前地震数据方位角校正方法的基础上，从方位角校正的角度入手，利用DMO和DMO^-1相结合的处理流程对地震数据进行规则化，以得到新的规则的观测系统下具有所需特定方位角的观测数据。此方法可用于解决叠前地震数据道插值、陆上地震勘探中不规则观测数据的规则化和海上地震勘探中的羽状现象校正等问题，可将已有地震观测数据转变为特定观测系统的叠前地震数据，以便于某些处理方法的应用。此方法原理简单并具有广泛的适用性。     

Solving regularized least squares with qualitatively controlled adaptive cross‐approximated matrices

The adaptive cross‐approximation (ACA) technique is applied to accelerating an inverse‐problem solver that estimates charge distribution on a dielectric spacer. The ACA generates an approximated system‐matrix that enables us to carry out high‐speed inverse calculation. We designed an approximation procedure based on ACA with some additional concepts, that is, (a) partitioning of matrix based on algebraic information, (b) approximation quality control based on matrix norms, and so on. The tested solver (LSQR for regularized least squares) with ACA demonstrates about 10 times faster performance than that without ACA. © 2007 Wiley Periodicals, Inc. Electr Eng Jpn, 159(3): 10–18, 2007; Published online in Wiley InterScience ( www.interscience.wiley.com ). DOI 10.1002/eej.20326     

基于正则总体最小二乘准则的OFDM系统信道估计算法

基于导频的最小二乘(LS)估计方法的优点是结构简单,容易实现,但对噪声和干扰敏感。通过对OFDM系统中由于信道的时变性产生的载波间干扰进行了分析,在此基础上提出正则总体最小二乘(TLS)信道估计算法。根据Tikhonov正则化原理,导出了总体最小二乘正则化解法的计算公式。与一般的LS信道估计相比,TLS方法同时考虑了信道噪声和信道时变特性,所以该算法不仅在一般的信道条件下具有良好的性能而且也适合快变信道条件下的OFDM系统。仿真结果表明,该算法具有很好的均方误差和误比特率性能。     

结构正则化多视图非负矩阵分解

现存非负矩阵分解（non-negative matrix factorization,NMF）研究多考虑单一视图分解数据,忽略了数据信息的全面性。此外,NMF限制其获取数据的内在几何结构。针对以上问题,提出一个结构正则化多视图非负矩阵分解算法（structure regularized multi-view nonnegative matrix factorization,SRMNMF）。首先,通过主成分分析来对数据进行全局结构的判别式学习;其次,利用流形学习来捕获数据的局部结构;然后,通过利用多视图数据的多样性和差异性来学习表征。模型提升了算法聚类的整体性能,更加有效地挖掘数据的结构信息。此外,采用高效的交替迭代算法优化目标函数得到最优的因子矩阵。在六个数据集上与现存的代表性方法比较,所提出的SRMNMF的准确率、NMI和Purity分别最大提高4.4%、6.1%和4.05%。     

基于方向场正则化的线描画生成算法

图像风格转化在计算机视觉领域广受关注, 其研究目标在于将输入图像利用计算机转化为具有某种特定艺术风格的图像. 线描画作为一种古老的画种, 它通过简单的线条勾勒物体的轮廓, 具有简约、抽象的风格. 本文提出一种基于方向场正则化的线描画生成算法, 该算法由4部分构成: 1)采用非局部平均滤波对输入图像进行预处理; 2)计算输入图像的方向场, 并基于自表示的思想对方向场进行Tikhonov正则化, 为了提高运算速度, 采用Sherman-Morrison-Woodbury公式来对正则化算法进行加速; 3)以正则方向场作为引导, 对预处理图像作高斯差分滤波; 4)根据人类视觉系统的非线性特点, 设计感知阈值(Perceptual thresholding)算法来对高斯差分滤波的结果进行阈值处理, 得到二值化的线描画图像. 仿真实验表明, 该算法可将输入图像转化为线条流畅且能有效表达输入图像主要信息的线描画图像.     

结构扩展的非负矩阵分解社区发现算法

提出结构扩展的非负矩阵分解社区发现算法(nonnegative matrix factorization with structure extension, NMF-SE),通过结构扩展,加强相邻节点结构相似性,提高节点间连接的稠密度,从而提高非负矩阵分解在社区发现中的表现。结构扩展过程使节点将自身结构以一定的比例传递给周围的节点,从而使相邻节点间能够得到对方的拓扑结构信息。该过程构造了新的特征矩阵,使非负矩阵分解(nonnegative matrix factorization, NMF)更好地适用于社区发现,在图正则化的半监督任务中能更好地融合先验信息。在人工网络和真实网络上进行试验验证的结果表明,NMF-SE算法有效提高了复杂网络社区发现的准确性。     

基于梯度正则化联合重构河流水质模型多项参数

为识别河流水质模型参数,首先采用Crank-Nicolson有限差分格式离散控制方程,其次通过附加终端观测值,设计梯度正则化算法重构河流水质模型的多项参数。结果表明,不管是顺直河流的常系数模型,还是天然的线性相关或者线性无关的变系数模型,梯度正则化方法均能快速有效识别模型参数;初值和测量误差对参数重构影响较小,充分证明梯度正则化算法是一种有效稳定地反演河流水质模型参数的方法。     

Total curvature variation fairing for medial axis regularization

We present a new fairing method for planar curves, which is particularly well suited for the regularization of the medial axis of a planar domain. It is based on the concept of total variation regularization. The original boundary (given as a closed B-spline curve or several such curves for multiply connected domains) is approximated by another curve that possesses a smaller number of curvature extrema. Consequently, the modified curve leads to a smaller number of branches of the medial axis. In order to compute the medial axis, we use the state-of-the-art algorithm from [1] which is based on arc spline approximation and a domain decomposition approach. We improve this algorithm by using a different decomposition strategy that allows to reduce the number of base cases from 13 to only 5. Moreover, the algorithm reduces the number of conic arcs in the output by approx. 50%.