WPSNR

对比CPU和GPU加速，pcl::cuda的使用教程，利用随机采样一致(RANSAC)去除地平面等例子。(Compare CPU and GPU acceleration, pcl::cuda tutorial, using random sampling consistency (RANSAC) to remove ground plane and other examples.)

LBP 一种描述子，常用于人脸识别，自己写的，包括实验程序和LBP函数(LBP descriptor)

说明：  图像处理案例，本程序用matlab实现基于阈值分割的车牌定位识别功能(Image processing case, this program uses MATLAB to achieve license plate location and recognition function based on threshold segmentation.)

使用极限学习机ELM算法进行遥感图像分类的源代码，很好的例子(Use extreme learning machine ELM remote sensing image classification algorithm source code, a good example)

提取特征点并匹配，有基于相关系数和最小二乘两种方法。(Extracting feature points and matching)

一个基于matlab开发的系统，实现对图像直方图的绘制、canny、sobel等几种不同算子的边缘检测(A system based on matlab development, drawing the image histogram, canny, sobel edge detection of several different operators can manually modulation threshold)

雾天图像增强算法，包括单尺度和多尺度的算法。(fog image enhancement)

帧间差法实现视频对象的分割，进行运动目标的检测小程序。(The frame difference method of video object segmentation, moving target detection procedures.)

该算法首先通过字典学习得到含噪图像的冗余字典，然后对相似的图像块进行聚类构成块群，并通过迭代收缩和L1正则化约束，对同类的图像块在字典上进行稀疏表示，以达到降噪的目的。实验结果表明，在常规的图像处理上，本文提出的算法能较好的保留图像的结构信息，与K-SVD和BM3D等现有的流行算法相比，具有更高的峰值信噪比（PSNR）(It firstly get the redundant dictionary of a noised image by dictionary learning.Then,the image patches are gathered according to their similarities.Meanwhile,the similar patches get sparse representation showed in dictionaries by iterative shrinkage and L1 regularization constraints and eventually the image is restored and noise is removed.The experimental results indicate that the proposed algorithm can well preserve the structure information of the common image with a higher Peak Signal to Noise Ratio(PNSR),compared with state-of-the-art algorithms,such as K-SVD and BM3D)

分形图像压缩 效果好 但是运行时间长 需要改进(Fractal image compression effective running a long time but need to improve)