magnify

基于机器视觉，控制三自由度运动控制台来调整显微镜位置，通过图像处理得到易拉罐的表面刻痕大小和长度。(The software control the motion platform with 3 d.o.f. to position the microscope and then test the notch of cans using image processing.)

好东西，下下来看就知道了好东西，下下来看就知道了(Good things, under the terms of the next good thing to know, under the circumstances of view to know)

图像的相关系数是衡量两幅图像的相关程度的，本代码直接通过选定区域计算出相关系数！(image correlation coefficient is a measure of two images of the code directly through selected regional correlation coefficient calculated!)

能够得到二维、三维voronoi图，修改后可以获得任意数量的voronoi图。(The two or thress-dimensional Voronoi diagram can be obtained, and any number of Voronoi diagrams can be obtained after modification.)

本程序是基于单幅图像的超分辨率算法，通过对输入的高分辨率图像进行下采样得 % 到原始低分辨率图像，对输入的低分辨率图像插值后，利用DWT将低分辨率图像分 % 解为不同的4个子带；同时直接对对分辨率图像进行SWT处理。由SWT得到的高频子 % 带来修正DWT得到的高频子带。最后，通过逆离散小波变换（IDWT）组合修正的高 % 频子带和低分辨率图像，得到输出的超分辨率图像。通过计算PSNR和SSIM评价结果。(This program is a super-resolution algorithm based on a single image, which is sampled from the high resolution image of the input. % to the original low resolution image, after interpolating the input low resolution image, the low resolution image is divided by DWT. The% solution is different 4 subbands, and the resolution image is processed by SWT. High frequency subsets obtained by SWT % brings the high frequency subband obtained by the modified DWT. Finally, the height of the combination of inverse discrete wavelet transform (IDWT) % frequency subband and low resolution image, and the output super-resolution image is obtained. The results were evaluated by calculating PSNR and SSIM.)

本人研究生阶段做的项目，线结构光视觉传感器的标定中的激光光条中心的提取。 1、通过JPG、PCX、GIF解码方法，来打开BMP、JPG、PCX、GIF格式图像。 2、大津展之法阈值变换，效果较好。 3、三种细化方法，Rosenfold细化、Hildith细化、Pavlidis细化。 4、基于灰度重心法的全分辨率的激光光条中心亚像素坐标的提取。(Calibration )

256色显示文件。在DOS环境下用256原色显示BMP格式的图片。(256-color display documents. In the DOS environment display with 256 color BMP format picture.)

knn k近邻算法，可选择欧式距离或者曼哈顿距离(knn k nearest neighbor, Euclidean distance or Manhattan can choose the distance)

说明：  a）针对合成纹理图像（共有4 个合成纹理图像，见文件夹： Texture_mosaic_data）中每一个像素提取纹理特征向量； b）利用聚类技术（推荐用k‐均值聚类，可以从网上查找原码）对特征向 量空间中的点进行聚类，类别数可根据图像中的实际纹理类数确定。最后把类 属标签映射成图像形式显示（如下图，其中b、d、f、h 为相应的基准分割图像）。(A) for synthetic texture images (there are 4 synthetic texture images in total, see the folder: Texture feature vector is extracted from every pixel in texture; B) use clustering technology (K-means clustering is recommended, and the original code can be found on the Internet) to search the feature direction The number of categories can be determined according to the actual number of texture classes in the image. Last class The generic label is mapped to an image (as shown in the figure below, where B, D, F, h are the corresponding reference segmentation images).)

a.牌照定位，定位图片中的牌照位置； b.牌照字符分割，把牌照中的字符分割出来； c.牌照字符识别，把分割好的字符进行识别，最终组成牌照号码。 (a. License Plate Location, location location of the picture of the license b. Character Segmentation license to license the characters split up c. license character recognition, to split a good character to identify the final composition of license number. )