analysefusion

使用PCA和SVD进行数据去噪，利用数据主要的特征向量进行数据恢复和重建(Using PCA and SVD for data denoising)

在IDL中运行，可以将多光谱、高光谱等多波段影像拆分为为单一波段影像(It works in IDL and it can split multibands into single band)

人脸识别matlab源代码，应用主分量分析（PCA）实现了人脸识别。(Face recognition matlab source code, application of principal component analysis (PCA) to achieve a face recognition.)

说明：  三维受激布里渊散射谱使用MATLAN仿真(Stimulated Brillouin Scattering)

基于灰度直方图的otsu（大津法）改进算法实现图片前景背景分离（matlab代码，实验报告）(Improved algorithm based on gray-scale histogram the otsu (Otsu method) background picture prospects separation (Matlab code, lab report))

故障分为很多种，该程序是利用小波包把故障特征提取出来，用于判断接下来出现的情况是把一种故障(The fault is divided into many, the program is using wavelet packet to extract the fault features out, used to judge the next case to a fault)

说明：  用曲波变换点边缘检测实现自动恢复受天气影响的图像(Qu Bo transformation point with edge detection for automatic restoration of images affected by the adverse weather)

程序将界面与数据处理分开，界面采用MFC的对话框模板，对注册、训练、识别和主界面分别建立一个类，通过消息处理函数与用户进行交互，利用多线程来实时显示图像。 数据处理又分为了CFaceAlign（人脸检测+几何归一化）、CLightPrep（光照归一化）、CFaceFeature（Gabor特征提取）、CSubspace（计算Fisherface子空间）四个类，还有一个类 CFaceMngr 负责管理界面与数据之间的交流。这样的程序结构使对界面和算法彼此独立，互不影响。 每个类的具体方法请见相应头文件中的注释。(The program will interface with the data processing separate interface using MFC dialog template, registration, training, recognition and the main interface, respectively, to create a class that interacts with the user through a message handler, using multiple threads to display images in real time.   Data processing is divided into a CFaceAlign (Face Detection+ geometric normalization), CLightPrep (illumination normalization), CFaceFeature (Gabor feature extraction), CSubspace (computing Fisherface subspace) four categories, as well as a class is responsible for managing CFaceMngr interface and data exchange between. Such a program structure allows for interfaces and algorithms are independent of each other. Specific methods of each class, see the comments in the appropriate header files.)

图像的采样 • 写程序：以2 为尺度，下采样图像 ，观察不同分辨率下图像的变化情况(a) • 写程序：把(a)中获得的图像，进行上采样处理，可以选择不同的平滑方 法，比如，邻近像素复制，或者线性插值等。讨论上采样后图像的质量变 化。找出相应的改进方法。

实现凸多边形排样，使用临界多边形NFP判断两个凸多边形是否相交,使用启发式算法提高效率(It s a good implement of convex polygon stock.)