IMM

幻方的构造及幻方实现的程序，对于其他双偶阶的幻方，只是在作‘*’标记时，应把4N阶方阵分为N个4阶方阵进行标记，数字填入仍按从左到右，自上而下填入(Magic Square and the magic square implementation procedures for other pairs of even order magic square, just as ' *' mark should be put into N 4N square matrix of order 4 squares labeled, digital still fill left to right, top to bottom fill)

基于matlab的图像小波融合测试程序，选取四副位图文件为对比(Matlab wavelet-based image fusion test program)

转：https://blog.csdn.net/jinshengtao/article/details/26278725 混合高斯背景建模是基于像素样本统计信息的背景表示方法，利用像素在较长时间内大量样本值的概率密度等统计信息(如模式数量、每个模式的均值和标准差)表示背景，然后使用统计差分(如3σ原则)进行目标像素判断，可以对复杂动态背景进行建模，计算量较大。 在混合高斯背景模型中，认为像素之间的颜色信息互不相关，对各像素点的处理都是相互独立的。对于视频图像中的每一个像素点，其值在序列图像中的变化可看作是不断产生像素值的随机过程，即用高斯分布来描述每个像素点的颜色呈现规律单模态(单峰)，多模态(多峰)(Gaussian mixture background modeling is a background representation method based on the statistical information of pixel samples. Statistical information such as the number of patterns, the mean and standard deviation of each pattern are used to represent the background. Statistical difference (such as 3_principle) is used to judge the target pixel. Complex dynamic background modeling has a large amount of computation. In the Gaussian mixture background model, it is considered that the color information between pixels is uncorrelated and the processing of each pixel is independent of each other. For each pixel in a video image, the change of its value in a sequential image can be seen as a random process that produces pixel values continuously, i.e. Gaussian distribution is used to describe the regularity of color rendering of each pixel in single mode (single peak) and multi-mode (multi-peak).)

对Lena图先进行jpeg压缩，再对压缩后的图像分别进行小波软阈值去噪和非抽样硬阈值去噪，比较两种去噪方案，非抽样硬阈值有更好的效果。这是根据一篇外文文献来进行编程的，最后有图像对比结果。(According to the paper of OPTIMAL WAVELET THRESHOLDING FOR VARIOUS CODING SCHEMES ,I firstly compresse the Lena picture using the jpeg method ,then respectivly denoise it using the wavelet soft-thresholding method and the undecimated DWT hard-thresholding method.Comparing the two method ,we can find the undecimated DWT hard-thresholding have a better result.)

基于纹理度量的图像分割，适用于遥感图像，利用到K-mean算法(texture measurement based on the image segmentation, applicable to remote sensing images, the use of K-mean algorithm)

自己编的分水岭算法的程序 大家随便看看吧(Own procedures for the watershed algorithm we just take a look at it)

本代码可以实现图像的维纳滤波，经过调试无错误，对于学习图像处理有较好帮助。(This code can achieve the image filtering, error-free through the debugging, better for learning image processing help.)

C++ 写的一些算法模板 ，比如Kruskal、KMP、欧几里德算法、最大匹配(匈牙利算法)、最小费用最大流、字符串的哈希方法、增广路算法、网络最大流_邻接表等……(C++ write some algorithms template, such as Kruskal, KMP, Euclid arithmetic, maximal matching (Hungary) is proposed, the minimum cost and maximum flow, string hash method, augmented road algorithm, network maximum flow _ adjlink etc......)

提出了一种针对复杂环境下的过人脸检测方法，首先在CbCgCr空间利用直接最小二乘法构建了对光照和复杂背景鲁棒性更好肤色聚类模型，实现了准确的肤色检测；然后针对人脸中姿态和表情变化，提出了基于Adaboost的多姿态人脸检测，精确的实现了人脸检测定位。(Extraordinary face detection method for complex environments, the first space CbCgCr ​ ​ direct least squares method to build a better robustness to light and complex background color clustering model to achieve accurate color detection forface, posture and facial expression changes, proposed the multi-pose face detection based on Adaboost, accurate face detection and location. )

数据库是由10个人的7种正面表情组成的213幅灰度图像，图像是以大小为256256的8位灰度级存储的，格式为.tiff型，平均每个人每种表情有2到4张。( the JAFFE database[11] which is composed of 213 images of female facial expression corresponding to 10 distinct subjects. Each image is stored at a resolution of 256×256 pixels and 8-bit gray level. Each subject in the database is represented with 7 categories of expression (angry, disgust, fear, neutral, sadness, happiness and surprise). )