2-D NONSEPARABLE SCALING FUNCTIONINTERPOLATION AND APPROXIMATION

The authors introduce nonseparable scaling function interpolation and show that its approximation can provide similar convergence properties as scalar wavelet system. Several equivalent statements of accuracy of nonseparable scaling function are also given. In the numerical experiments, it appears that nonseparable scaling function interpolation has better convergence results than scalar wavelet systems in some cases.

Fractal characteristics investigation on electromagnetic scattering from 2-D Weierstrass fractal dielectric rough surface

A normalized two-dimensional band-limited Weierstrass fractal function is used for modelling the dielectric rough surface. An analytic solution of the scattered field is derived based on the Kirchhoff approximation. The variance of scattering intensity is presented to study the fractal characteristics through theoretical analysis and numerical calculations. The important conclusion is obtained that the diffracted envelope slopes of scattering pattern can be approximated as a slope of linear equation. This conclusion will be applicable for solving the inverse problem of reconstructing rough surface and remote sensing.

Electronic Absorption Spectra and Third-Order Nonlinear Optical Property of Dinaphtho[2,3-b:2’,3’-d]Thiophene-5,7,12,13- Tetraone (DNTTRA) and Its Phenyldiazenyl Derivatives: DFT Calculations

Third-order nonlinear optical (NLO) materials have broad application prospects in high-density data storage, optical computer, modern laser technology, and other high-tech industries. The structures and frequencies of Dinaphtho[2,3-b:2’,3’-d]thiophene-5,7,12,13-tetraone (DNTTRA) and its 36 derivatives containing azobenzene were calculated by using density functional theory B3LYP and M06-2X methods at 6-311++g(d, p) level, respectively. Besides, the atomic charges of natural bond orbitals (NBO) were analyzed. The frontier orbitals and electron absorption spectra of A-G5 molecule were calculated by TD-DFT (TD-B3LYP/6-311++g(d, p) and TD-M06-2X/6-311++g(d, p)). The NLO properties were calculated by effective finite field FF method and self-compiled program. The results show that 36 molecules of these six series are D-π-A-π-D structures. The third-order NLO coefficients γ (second-order hyperpolarizability) of the D series molecules are the largest among the six series, reaching 107 atomic units (10-33 esu) of order of magnitude, showing good third-order NLO properties. Last, the third-order NLO properties of the azobenzene ring can be improved by introducing strong electron donor groups (e.g. -N(CH3)2 or -NHCH3) in the azobenzene ring, so that the third-order NLO materials with good performance can be obtained.

An Extended Subequation Rational Expansion Method and Solutions of (2+1)-Dimensional Cubic Nonlinear Schr(?)dinger Equation

An extended subequation rational expansion method is presented and used to construct some exact,analyt-ical solutions of the (2+1)-dimensional cubic nonlinear Schrdinger equation.From our results,many known solutionsof the (2+1)-dimensional cubic nonlinear Schrdinger equation can be recovered by means of some suitable selections ofthe arbitrary functions and arbitrary constants.With computer simulation,the properties of new non-travelling waveand coefficient function's soliton-like solutions,and elliptic solutions are demonstrated by some plots.

基于递归Gabor滤波的指纹图像处理方法的研究

根据一维递归Gabor滤波算法设计了二维RGF(Recursive Gabor Filtering)算法,通过实验论证了RGF不仅与Gabor滤波等价,而且速度更快。在此基础上提出了RGF方法代替传统的Gabor滤波器方法应用到指纹特征提取,并用实验论证了参数的选择。实验结果显示该方法的有效性,具有很高的实用价值。

基于LVQ神经网络的植物种类识别

提出一种基于学习矢量量化(LVQ)神经网络的计算机植物种类识别新方法.使用2-D不变矩、多尺度2-D Gabor滤波器等多种方法分别提取了叶片的几何特征和纹理特征,应用LVQ神经网络识别植物种类.实验结果表明,该方法对植物种类的识别效率较高.

基于移不变全方向角提升的遥感图像降噪

在分析遥感图像结构特征及其与噪声之间主要区别的基础上,利用图像信号的方向信息,提出基于移不变全方向角提升小波(TI-OL)抑制遥感图像噪声的方法。该方法在方向提升小波变换的基础上并利用循环平移,Gabor小波滤波器和图像旋转技术改进了方向提升小波在图像去噪过程中存在的三个弊端:缺乏移不变性质,图像局部方向信息判方法断缺乏噪声鲁棒性和变换方向分布有限。消除去噪结果中的吉布斯效应,提高图像方向信息判断的准确性并保证图像纹理方向始终落在方向提升能最优表示的方向区间内。试验结果证明所提方法在处理遥感图像的过程中能在去噪的同时保留图像的细节和边缘信息,对遥感图像中的边缘信息如道路和桥梁有较好的刻画性能,较传统方法去噪性能(PSNR)和主观视觉效果(SSIM)均有较大提高。

基于纹理检测器的多频道纹理分析

按照２ＤＧａｂｏｒ函数的特点和视觉机制，提出了用来捕捉纹理基元的纹理检测器函数．基于纹理检测器和多频道分解理论，提出了影像纹理的多频道分析方法，并按照神经动力学的侧抑制和端点抑制等理论，实现了对多频道纹理特征的融合，为改进现存多频道滤波方法提供了良好途径．实验表明，所提出的方法可有效用于具有相似模式大小的纹理分析．

A NOVEL METHOD ON DESIGNING BISPECTRAL WINDOWS

A novel method on designing bispectral windows is proposed in this letter. Compound functions based on a function with binary quadric form, starting from the symmetry characteristic of three-order moments are used as the 2-D windowfunctions. Two approaches on how to find the expressions of the compound functions are discussed in detail. One is to approximate the compound function after being Taylor expanded under the Minimum Mean Square Error (MMSE) criteria. Another is to compound the hyperbolic secant function and the binary quadric function directly. According to theoretical analysis, the first type new windows have been proved as slightly better than the conventional ones, the second type new windows are much better than the conventional ones, and the bispectral estimation mean square error approximates to 0.

Two Dimensional Spatial Independent Component Analysis and Its Application in fMRI Data Process

One important application of independent component analysis （ICA） is in image processing. A two dimensional （2-D） composite ICA algorithm framework for 2-D image independent component analysis （2-D ICA） is proposed. The 2-D nature of the algorithm provides it an advantage of circumventing the roundabout transforming procedures between two dimensional （2-D） image deta and one-dimensional （l-D） signal. Moreover the combination of the Newton （fixed-point algorithm） and natural gradient algorithms in this composite algorithm increases its efficiency and robustness. The convincing results of a successful example in functional magnetic resonance imaging （fMRI） show the potential application of composite 2-D ICA in the brain activity detection.