qicanshu

Camshift算法，即"Continuously Adaptive Mean-Shift"算法，是一种自适应的运动跟踪算法。它主要通过视频图像中运动物体的颜色信息来达到跟踪目的。(Camshift algorithm, namely " Continuously Adaptive Mean-Shift" algorithm is an adaptive motion tracking algorithm. It is primarily the color information of the object' s movement through the video images to achieve tracking purposes.)

灰度图生成程序 可以生成BMP/JPG格式的灰度图 大家可以随便使用下载(Grayscale generator can generate BMP/JPG format, you can easily use the download grayscale)

生成相机标定所用的方格图，用于相机标定使用(Generates the grid chart used to calibrate the camera)

这本是用VC++来实现图像处理的书藉,对于学习图像处理的来说是非常有用的.(This is VC++ to realize image processing books, for learning image processing is very useful.)

实现快速的 TV去噪，里面有相关的几篇论文 有全面的源代码(de-noising by TV)

中国科学院长春光学精密机械与物理研究所 基于外极线约束条件的视频判读方法(Changchun Institute of Optics, Fine Mechanics and Physics epipolar constraint based on the video interpretation method)

指纹图像因设备、图像增强处理等原因造成的纹线断裂会产生伪特征点,会对指纹识别产生负面影响。在结 合传统方法优点的同时,针对间距大、方向偏移小和间距大、方向偏移大的断裂情况,本文设计出窗口制约法和改进的 窗口制约法2 种不同的修复方法。实验结果表明,该方法使修复更加全面、完整,提高了伪特征点去除率,易于编程实 现,对细化指纹图像的断裂有很好的修复效果。(The broken ridge fingerprint image processing equipment, image enhancement and other causes will generate pseudo-feature points to be a negative impact on fingerprint identification. While combining the traditional method has the advantage, for the spacing, direction offset small spacing, direction offset large fracture situation, this design window constraints window method and the improved two kinds of constraints, different repair methods. Experimental results show the method to fix a more comprehensive and complete removal of pseudo-feature points, easy to program, well thinned fingerprint image fracture repair.)

一种基于non-sample的纹理合成demo，程序性能还有待优化，效果还不错.(texture synthesis demo.)

Pyramid methods in image processing

三维激光点云配准是点云三维建模的关键问题之一。经典的 ICP 算法对点云初始位置要求较高且配准 效率较低，提出了一种改进的 ICP 点云配准算法。该算法首先利用主成分分析法实现点云的初始配准，获得较好 的点云初始位置，然后在经典 ICP 算法的基础上，采用 k － d tree 结构实现加速搜索，并利用方向向量夹角阈值去除 错误点对，提高算法的效率。实验表明，本算法流程在保证配准精度的前提下，显著提高了配准效率。 (Three-dimensional laser point cloud registration is one of the key three-dimensional point cloud model. High classical ICP algorithm to the initial position of the point cloud registration requirements and low efficiency, proposed an improved ICP point cloud registration algorithm. Firstly, the use of principal component analysis of the initial point cloud registration, get a better initial position of the point cloud, then the basis of classical ICP algorithm using k- d tree structure to achieve speed up the search, and using the direction vector angle the removal of the threshold point error and improve the efficiency of the algorithm. Experiments show that the algorithm processes to ensure the accuracy of registration under the premise, significantly improve the efficiency of registration.)