cavi

品质因子可调的离散小波变换工具箱，可通过参数灵活控制小波函数的频域窗宽(Discrete wavelet transform toolbox with quality factor adjustable, flexible control parameters through the wavelet function in the frequency domain window width)

SVM 用于基于块划分特征提取的图像分类源码(SVM for feature extraction block-based image classification into source)

图像质量评价代码，用于评估两副图片的相似度(Image quality evaluation)

说明：  经典的基于非局部思想的图形图像去噪模型，也即非局部滤波，基于matlab实现(The classical image denoising model based on nonlocal idea, namely non local filtering, is realized based on MATLAB)

用于实现高密度人群聚合检测 包含：边缘检测、灰度共生矩阵计算、SVM特征分量的分类(For achieving high-density population polymerization detector comprises: edge detection, gray level co-occurrence matrix calculation, classification SVM characteristic component)

实现了立体匹配的SAD算法计算视差图，结果比较理想(Using the SAD algorithm to calculate the disparity map of stereo matching, the result is good.)

针对CT 医学图像和MR I 医学图像成像特点, 提出了基于快速整数提升小波变换的融合方法。在CT 和 MR I 两幅医学图像配准的前提下, 利用提升小波变换把图像分解成低频和高频子图像, 对于小波变换后的高频 子图像, 选择区域标准差大的作为融合后的子图像 对于低频子图像, 采用加权融合, 最后进行小波逆变换, 得到 融合后的图像, 并对融合后图像用信息熵、平均梯度、相关系数的指标进行评价。实验结果表明, 基于快速整数提 升小波变换融合中, 小波高低频系数采用不同的规则能够取得更好的融合效果, 其轮廓清晰。该算法能够提升融 合后的医学图像信息量, 同时有效地保护图像的细节信息, 在执行时间和图像质量上均优于普通小波算法。(A imed at the characterist ics of CT andMR Imedical images, a new image fu sion al2 go rithm based on the fast in t lif t ing w avelet t ran sfo rm s is p ropo sed. Two o riginal images are decompo sed by the fast in t lif t ing w avelet t ran sfo rm s. The fu sion ru le based on the max imum standard deviat ion value variance is u sed to fu se the h igh f requency sub2image, w h ile a w eigh t2 ed average fu sed ru le is app lied by coeff icien t s of the low f requency. F inally, the fu sion image is recon st ructed fo r perfo rm ing inverse fast in t lif t ing w avelet t ran sfo rm s. The fu sion image is evaluated by en t ropy, average gradien t, and co rrelat ion coeff icien t s. Experimen tal resu lt show s that the fu sion image hasmo re info rmat ion than o riginal images and imp roves the quali2 ty of o riginal images. M eanw h ile, the fu sion image p ro tect s detail characterist ics of the image, thu s the real2t ime p rocess and image qualit ies are w ell than tha)

说明：  盲迭代反卷积算法，可以在不求PS的情况下恢复图像(Iterative blind deconvolution algorithm, can not seek restoration of the PS images)

基于opencv的MHT的多目标跟踪程序(Based on the MHT multi-target tracking program)

说明：  实现了双线性插值、三次样条插值、B样条插值和最大阶次优化差值等传统多项式图像插值办法(Traditional polynomial image interpolation methods such as bilinear interpolation, cubic spline interpolation, B spline interpolation and maximum order optimization difference are realized.)