optiflow

手持式摄像机在使用时常常会受到使用者有意无意抖动的影响,从而影响成像效果，造成录制视频的不稳定及跳动问题，尤其是在使用者在一场景中特写或者跟踪某一具体目标时，使用者通常不能准确定位到或者估计出运动目标的位置，从而造成目标在视频中位置的不稳定，造成视频的主观效果变得不理想。 为了解决这一问题，我们需要设计一种算法来识别这种无意义的运动并设法通过补偿的方式来使得场景中的目标物体保持位置稳定的状态。 手持式摄像机捕获的视频通常都会受到抖动的影响，这严重的影响视频的主观效果。 这里提出的算法可以可靠的用于数字视频的去抖动。这个算法通过识别这些意外的抖动并且利用运动补偿的方法来获得一个较好的视频输出。 这个系统可以分为三个模块：（1）运动估计模块；（2）抖动识别模块；（3）运动补偿模块。 注：代码在matlab 7.7.0(R2008b)运行通过 [1] F Vella, A Castorina, M Mancuso. Digital image stabilization by adaptive block motion vectors filtering, IEEE Transactions on Consumer Electronics, 2002(Handheld cameras are often used by the user intentionally or unintentionally, the influence of jitter affect the imaging results, resulting in instability and beat the problem of recording video, and especially in the user in a scene close-up or tracking a specific target,users often can not accurately navigate to estimate the location of the moving target, resulting in the subjective effects of target location of the instability in the video, resulting in video not ideal. In order to solve this problem, we need to design an algorithm to identify such a meaningless exercise and try to make the object in the scene to keep the position stable state by way of compensation. The handheld camera captured video usually subject to the influence of jitter, the subjective effect of which seriously affect the video. The algorithm proposed here can be reliably used for digital video to jitter. This algorithm by identifying the jitter of these accidents and the use of motion compensation m)

分为行人检测，特征提取，行人跟踪三步，其中最主要的代码也是这个代码的中心是行人跟踪这一部分，我采用的是kalman滤波器进行行人下一帧位置的预测，可以很好的提高行人跟踪的鲁棒性，跟踪结果比较准确。(Into pedestrian detection, feature extraction, pedestrian tracking steps, which is the center of the main code of the code is pedestrian tracking this section, I kalman filter is used to predict the position of the next frame pedestrians, can be a good improvement pedestrian tracking robustness, tracking results more accurate.)

使用matlab实现的,基础单变量以及多变量梯度下降算法,和信号分类以及信号归一化部分算法，是机器学习领域最基本的代码，希望大家多多交流(matlab implementation of basic univariate and multivariate gradient descent and signal classification algorithms, machine learning basic code, hope that more exchanges)

用于求解凸优化问题，恢复低秩密度矩阵问题(this matlab Algorithm can be used for the convey optimal problen, and also for the low-rank density matrix reconstruction)

一个全面的图像处理平台，基本上具有教课书上所说的所有图像处理算法，并附带一些关键算法！()

对图像进行简单处理，对光斑圆拟合，求圆心与半径(Simple processing of the image, fitting the spot circle, finding the center of the circle and radius)

use js and css show image change effect

各种非监督分类遥感算法的IDL代码，包括K均值，ISO聚类(Remote sensing of various non-supervised classification algorithms IDL code, including K-means, ISO clustering)

说明：  kl变换. KL变换的流程如下： （1)获取多幅图像的统计特性：图像之间的协方差 （2)获取协方差的特征值和特征向量，从而获取正交核:特征矩阵 （3)图像对应象素点对各个特征向量进行投影运算。(kl transformation)

说明：  基于四元数的matlab处理工具包，利用四元数实现了很多图像处理的功能，并提供了api供大家使用(Quaternion-based processing tool kit of matlab, the use of quaternions to achieve a lot of image processing functions, and to provide for your use of the api)