EyeTracking-with-OpenCV--

Retinex理论的基础理论是物体的颜色是由物体对长波（红色）、中波（绿色）、短波（蓝色）光线的反射能力来决定的，而不是由反射光强度的绝对值来决定的，物体的色彩不受光照非均匀性的影响，具有一致性，即retinex是以色感一致性（颜色恒常性）为基础的。不同于传统的线性、非线性的只能增强图像某一类特征的方法，Retinex可以在动态范围压缩、边缘增强和颜色恒常三个方面打到平衡，因此可以对各种不同类型的图像进行自适应的增强。(The basic theory of Retinex theory is the color of the object by object for the long wavelength (red), medium (green), short wave (blue) light reflection ability to decide, rather than by the intensity of reflected light of the absolute value to decide, the color of the object under illumination nonuniformity effects, consistent with the Retinex, that is based on color constancy (color constancy based). Different from the traditional linear, nonlinear image enhancement method of only one kind of features, Retinex can be in dynamic range compression, edge enhancement and three aspects of color constancy hit the balance, thus enhancing adaptive to a variety of different types of images can be.)

图像识别的小软件，识别率还算凑合吧，大家共同学习(Small image recognition software, the recognition rate still make it, we learn together )

中值滤波，取出奇数个数据进行排序，用排序后的中值取代要处理的数据即可(median filtering.rar)

《基于计算机视觉的普适性三维重建方法研究》源代码，实现了一种将二维图像重建成三维场景的方法。(Source code of <General Sense Three-Dimension Reconstruction Research Based on Computer Vision>, a technique which can convert 2D images into 3D sence.)

（2001何斌+光盘）Visual c++数字图像处理（第二版），使用VC进行数字图像处理的相关资料，十分实用！((2001 Bin+ disc the) Visual c++ digital image processing (second edition), digital image processing using VC, very practical!)

该算法可以应用于计算机视觉的物体识别以及3D重构中。(The algorithm can be applied to object recognition in computer vision and 3D Reconstruction.)

傅里叶变换：采用VC++编写，实现二维图像的傅里叶变换(Fourier Transform: The VC++ prepared to achieve the two-dimensional Fourier transform of the image)

This is a Scaling function definition Image cutoff ratio for H = 3

利用Artoolkit实现了虚拟现实的算法，VS2008下测试通过(Use Artoolkit algorithm to achieve a virtual reality, VS2008 under test)

一种简单高效地基于压缩感知的跟踪算法。首先利用符合压缩感知RIP条件的随机感知矩对多尺度图像特征进行降维，然后在降维后的特征上采用简单的朴素贝叶斯分类器进行分类。该跟踪算法非常简单，但是实验结果很鲁棒，速度大概能到达40帧/秒(A simple and efficient tracking algorithm based on compressed sensing. Firstly, with the random sensing matrix compressed sensing RIP conditions for multi-scale image feature dimension reduction, and then use the naive Bias classifier simple classification in the feature reduction after the. The tracking algorithm is very simple, but the results are robust, speed can reach 40 frames per second)