img_resample

说明：  生成手写数字图片，一共提供了三种算法，example_1和example_2训练GAN来生成手写数字图片，搭建了一个简单的GAN的网络结构，example_3难度较大(There are three algorithms to generate handwritten digital pictures. Example_1 and example_2 train GAN to generate handwritten digital pictures, and build a simple network structure of GAN. Example_3 is more difficult.)

说明：  通过生成秘钥得到一串具有一定长度的扩频码，将原始消息序列以扩频码的长度过采样产生过采样消息序列，并将扩频码以原始消息序列的长度多次重复成伪随机序列，再令过采样消息序列与伪随机序列异或就可以生成扩频消息序列。在频率域下，将扩频消息序列以一定的嵌入强度自适应地嵌入原始图像，得到的图像具有较强的鲁棒性，比起直接嵌入，扩频水印在受到攻击后恢复出的水印信息误码率要低得多。并且扩频码长度越长，容错率就越高，鲁棒性就越好。(A sequence of spread spectrum codes with a certain length is obtained by generating a secret key. The oversampling message sequence is generated by oversampling the original message sequence with the length of the spread spectrum code, and the spread spectrum code is repeated into a pseudo-random sequence with the length of the original message sequence for many times, and then the spread spectrum message sequence can be generated by making the difference between the oversampling message sequence and the pseudo-random sequence. In the frequency domain, the spread spectrum message sequence is adaptively embedded into the original image with a certain embedding strength, and the resulting image has strong robustness. Compared with direct embedding, the error rate of the watermark information recovered from the attack of spread spectrum watermark is much lower. And the longer the spreading code length, the higher the fault tolerance rate and the better the robustness.)

一个比较不错的水平集改进算法，内附说明文档。(A relatively good level set to improve the algorithm, containing documentation.)

二维码扶正,对不同光照图片都有较好结果，涉及几种不同的二值化方法，采用霍夫检测直线和双线性变换扶正二维码，内含的7张图片都能扶正(Two-dimensional code righting, have a better picture of the different light result, involving several different binarization method using Hough transform to detect linear and bilinear righting two-dimensional code, embedded image 7 can be righting)

给予稀疏编码的原理将10幅自然图像进行系数分解成为图像基。(Given to the principle of sparse coding will be 10 coefficients of natural images into separate image-based.)

一个基于脉冲耦合神经网络的高斯噪声滤波程序，包括源码和实验结果(pulse coupling based on a neural network Gaussian noise filtering procedures, including source and experimental results)

图像匹配的实现，其中match.m：测试程序，sift.m :尺度不变特征变换（SIFT算法）的核心算法程序，appendimages.m: 该函数创建一个新的图像分别包含两个匹配的图像和他们之间的匹配对的连接直线，可以运行。 (The realization of image matching, which match.m: test program, sift.m: scale-invariant feature transform (SIFT algorithm) of the core algorithm, appendimages.m: This function creates a new image each consists of two images and matching match between them on the connection line, you can run.)

说明：  matlab图像滤波去噪分析及其应用 双线性滤波、Kirsch滤波、超限邻域滤波、逆滤波、双边滤波、同态滤波、小波滤波、六抽头滤波、约束最小平方滤波、非线性复扩散滤波、Lee滤波、Gabor滤波，、Wiener滤波、Kuwahara滤波、Beltrami流滤波、Lucy?Richardson滤波、NoLocalMeans滤波等研究内容。(Analysis and application of MATLAB image filtering and denoising Bilinear filtering, Kirsch filtering, transfinite neighborhood filtering, inverse filtering, bilateral filtering, homomorphic filtering, wavelet filtering, six tap filtering, constrained least square filtering, nonlinear complex diffusion filtering, Lee filtering, Gabor filtering, Wiener filtering, Kuwahara filtering, Beltrami flow filtering, Lucy Richardson filtering, nolocalmeans filtering, etc.)

压缩感知图像处理用于SAR 经典的压缩感知教科书的源代码(compressed sensing SAT)

使用极限学习机ELM算法进行遥感图像分类的源代码，很好的例子(Use extreme learning machine ELM remote sensing image classification algorithm source code, a good example)