muller

说明：  利用全卷积神经网络，实现图像的语义分割，基于tensorflow的keras可以直接运行(Using the full convolutional neural network to achieve semantic segmentation of images, keras based on tensorflow can be run directly)

sift图像处理的一个测试的代码，他是用C++写的 用到了opencv中的一些函数(sift image processing as a test of the code, he wrote with C++ uses some of the opencv function)

细化。通过MATLAB对图像进行细化提取骨架处理，从而方便进行下一分析(Refinement zs appises the Zhang-Suen skeletonization algorithm to image IM. IM must be binary.)

说明：  对灰度图像和二值图像进行去噪，是一种形态学滤波器。(denoiseing for the gray image and binary image)

利用curvelet波将不同聚焦图像进行分解，然后利用局部能量最大的方法将图像合成，得到清晰的图像、(Curvelet waves using different focus image decomposition, and then make use of local energy will be the largest image synthesis methods have been given clear images,)

基于Matlab的结合Laplacian Pyramid 和反锐化掩膜的改进法的图像增强算法(Matlab based on a combination of Laplacian Pyramid and unsharp mask to improve the method of image enhancement algorithm)

通过自定义数学运算，在图像处理中做直方图均衡化的matlab实现(Through custom mathematical operation, do the histogram equalization of matlab in image processing)

纹理模式的LBP特征表示及分类原理：就是通过定义一个模板，对于原图像中的每个点，使用该模板进行操作，模板中心点近邻区域对应的图像象素值与中心位置对应的象素值进行比较，且对于不同位置的比较结果赋予不同的权值，从而，变换图像中对应点的值就由模板对应各位置的比较结果之权和表示。(LBP texture pattern feature representation and classification theory: that by defining a template for the original image of each point, use the template to operate, the template center of neighboring regions corresponding to the center of the image pixel values corresponding to the pixel value comparison, and the comparison results for different locations given different weights, and thus, changing the value of the image corresponding points on the template corresponding to each position by the comparison of the results of the right to and express.)

Mallat算法和Atrous算法实现任意一维信号和二维图像的分解与重构。(Mallat algorithm and Atrous algorithm to achieve the arbitrary one-dimensional signals and two-dimensional image decomposition and reconstruction.)

使用Tensoflow实现图卷积神经网络模型(Use Tensoflow Implementation Figure convolution neural network model)