matlab-3d-reconstruction

基于贝叶斯最小错误率算法的崇明岛遥感图像分类程序。并且有饼状图显示各类地物比例(Minimum error rate based on Bayesian algorithm for Chongming Island, the remote sensing image classification procedures. And the pie chart below shows the proportion of various types of features)

该程序初始时使用背景差分法自动检测目标，当目标完全进入图像中后，改用meanshift方法对目标进行跟踪。(the program use background subtraction to detect the target first,when the target fully enter the screen,change to meanshift method to track the target.)

视觉注意力模型，寻找感兴趣区域，模仿人眼找出最感兴趣的区域，再找出第二感兴趣的区域，以此类推(Visual attention model to find the region of interest, to imitate the human eye to identify the most interesting region, and then find the second region of interest, and so on)

3种分别基于选择、交叉、模拟退火的混合粒子群算法，matlab编程。(Were based on three kinds of selection, crossover, hybrid particle swarm simulated annealing algorithm, matlab programming.)

HBT实验，这个实验用来测量恒星的角直径。这个算法可以对实验进行仿真(HBT experiment, this experiment is used to measure the angular diameter of stars. This algorithm can simulate the experiment)

　对经典的二值化方法Ostu 算法和Bernsen 算法中存在的缺点进行了分析, 提出一种结合Canny 算子的图像 二值化方法.该方法综合考虑了边缘信息和灰度信息, 通过边缘附近种子点在高阈值二值化图像中的填充和低阈值 图像对它的修补而得到二值化结果图像, 较好地解决了经典二值化方法中存在的抗噪能力差、边缘粗糙、伪影现象 等缺点.实验结果证明, 该方法能够较好地解决低对比度图像和目标像素灰度不均匀图像的二值化问题(The disadvantage of the classical binary method Ostu Bernsen algorithm and algorithms exist were analyzed, one binding Canny operator image binarization method. This method considers the edge information and gray information, through seed near the edge point padding and a low threshold image to its repair in a high threshold binary image to obtain a binary result image, it solves the noise immunity classical binary method exists difference, rough edges, artifacts phenomenon and other shortcomings. the experimental results show that this method can solve the low-contrast image and the target image pixel gray uneven binary problem)

说明：  这是用来做直方图匹配的工具，测试过可以实现(Histogram Matching Tool)

从原始图像产生噪声图像，使用db1执行2层小波，使用wpbmpen进行全阈值选择，采用硬阈值和软阈值处理方式,保存低频,进行图像降噪，并绘制各分解系数图像。(Generated noise image from the original image, using db1 wavelet perform two layers using wpbmpen full-threshold selection, using hard threshold and soft threshold approach, saving low frequency, image noise reduction, and draw the decomposition coefficients images.)

林健写的Matlab在图像处理与目标识别方面的应用实验，包括染色体识别与统计、汽车牌照定位与字符识别、基于BP神经网络识别字符的简单实验。 (Lam Kin Writing at the Matlab image processing and target recognition application experiments, including chromosome identification and statistics, vehicle license positioning and character recognition, based on BP neural network easy to identify characters in the experiment.)

噪声估计法，先定位噪声点，再实行去噪处理(Noise estimation method, first locate the noise points, then implement denoising.)