Assignment3

说明：  简单的graph cut 算法，初学者适合(Simple graph cut algorithm for beginners)

基于mmdetection框架的深度学习目标检测框架(deep learning object detection based on mmdetection)

该程序是帧差法和背景法相结合的一种方法，非常新颖(The program is the frame difference method and the background of a combined approach, a very novel)

基于大津法和迭代法的图像二值化程序，相当于利用该方法进行了图像的分割，源码已调试，很好使！(Based on the Otsu method and iterative method of image binarization procedure is equivalent to using the method of image segmentation, source code has been debugging, very good to make!)

波原子变换代码，一种新的类方向小波变换，更好的表示图像(Wave atom transform code, the direction of a new class of wavelet transform, and better representation of the image)

基于混沌置乱的DCT域灰度级盲水印算法 需要的可以看下~(Based on Chaotic Scrambling DCT-domain blind watermarking algorithm for gray-level needs can facie ~)

双目摄像机标定程序，给出摄像机内外参数，并完成三维立体重建。(Binocular camera calibration procedures, given internal and external camera parameters, and complete the three-dimensional reconstruction.)

采用色温估计法做的白平衡操作，之前是打算用到水下处理的，当然这个算法和PS的白平衡比有点LOW，可是这个能够matlab，有一定用途(The white balance operation using the color temperature estimation method was intended to be used underwater. Of course, the white balance ratio of this algorithm and PS is a little LOW, but this can be used in matlab.)

此程序为用于主成分分析的matlab程序,可以输出贡献率及画出二维散点图(This program for principal component analysis of matlab, can output contribution rate and draw a 2 d scatter plot)

针对电气设备同一场景的红外与可见光图像间一致特征难以提取和匹配的问题，提出了一种基于斜率一致性的配准方法。首先通过数学形态学方法分别提取红外与可见光图像的边缘，得到粗边缘图像；然后通过SURF算法提取两幅边缘图像的特征点，根据正确的匹配点对之间斜率一致性的先验知识，进行特征点匹配；最后通过最小二乘法求得仿射变换模型参数并实现两幅图像的配准。实验结果表明，该方法有效提高了匹配点对的正确率，能够对电气设备红外和可见光图像实现高精度的配准。(In this paper, a method based on slope consistency is proposed to solve the problem that the uniform features of infrared and visible images are difficult to extract and match. Firstly, the edges of infrared and visible images were extracted by mathematical morphology, and the rough edge images were obtained. Then, the feature points of two edge images were extracted by SURF algorithm, and the priori knowledge of slope consistency was obtained according to the correct matching point, and the feature points were matched. Finally, the model parameters of affine transformation are obtained by least square method and the registration of two images is realized. The experimental results show that this method can effectively improve the accuracy of matching points, and can realize high precision registration of infrared and visible images of electrical equipment.)