预处理

说明：  用分水岭算法进行图像分割，借鉴形态学理论，将一幅图看成拓扑性地图，其中灰度值对应地形高度。(The watershed algorithm is used for image segmentation, and a map is regarded as a topological map, in which the gray value corresponds to the terrain height.)

说明：  HDMI接口定义，包括a，b，c，d，包括各种接口的详细介绍(HDMI interface definition, including a, b, c, d, including detailed introduction of various interfaces)

利用MATLAB进行数值计算，得出HE模的色散曲线。(The dispersion curve of HE mode is obtained by numerical calculation by MATLAB.)

说明：  OpenGL的示例文件。上课时介绍用的示例文件，内含exe文件可直接运行。(OpenGL's sample file. Sample files for introduction in class, including exe files, can be run directly.)

TBD，检测前跟踪算法的仿真程序，比较多的文件，自己区分(TBD, tracking algorithm pre-test simulation procedures, a relatively large number of documents, to distinguish between their own)

说明：  由于只是利用图像的灰度信息, SIFT 算法不能很好地区分形状相似但颜色不同的物体。针对这一问题, 提出了一种基于彩 色的SIFT 特征点提取算法, 并着重分析了多种彩色模型对算法性能的影响。这种算法也是在图像的灰度尺度空间上检测特征点, 但其特征向量由各描述子子区域的彩色模型分量的均值组成并在原始的彩色图像上进行计算。(SIFT)

说明：  运行小波变换对图像进行处理分析，包括去噪纹理提取等(Running wavelet transform to process and analyze image, including denoising texture extraction, etc.)

通过使用LLE PCA两种方法相结合的思路，提高人脸识别的分别率(LLE PCA by using the idea of combining two methods to improve face recognition rate difference)

说明：  虚拟内窥镜关键技术研究 虚拟内窥镜技术（Virtual Endoscopy）是虚拟现实技术在现代医学中 的应用。它利用医学影像作为原始数据，融合图像处理、计算机图形学、科 学计算可视化、虚拟现实技术，模拟传统光学内窥镜的一种技术。它克服了 传统光学内窥镜需把内窥镜体插入人体内的缺点，是一种完全无接触式的检 查方法。虚拟内窥镜的研究旨在为医生提供诊断依据，还可应用于辅助诊断、 手术规划、实现手术的精确定位和医务人员的培训等。(Research on key technologies of virtual endoscopy virtual endoscopy (Virtual Endoscopy) is a virtual reality technology in modern medicine. It has made use of medical imaging as an original data, the integration of image processing, computer graphics, scientific visualization, virtual reality technology to simulate a conventional optical endoscope technology. It overcomes the traditional optical endoscope to be inserted into the endoscope body of the shortcomings of the human body is a no-contact inspection method. Virtual Endoscopy of the study was to provide diagnosis for doctors, but also applies to diagnosis, surgical planning, to achieve precise positioning of surgical and medical personnel training.)

DensePose用深度学习把2D图像坐标映射到3D人体表面上，再加上以每秒多帧的速度处理密集坐标，最后实现动态人物的精确定位和姿态估计。该技术集目标检测、姿态估计、目标部分/实例分割等多种计算机视觉任务于一身的一个综合问题。(DensePost maps 2D image coordinates to 3D human body surface by in-depth learning, and processes dense coordinates at the speed of multiple frames per second. Finally, it realizes precise positioning and attitude estimation of dynamic characters. This technology integrates many kinds of computer vision tasks, such as target detection, attitude estimation, target part/instance segmentation, etc.)