gaosi

最简单的进行两图像的加减和异或，同或，运算。(The simplest operation is to add and subtract two images, and to add and subtract two images.)

说明：  经典图像处理方法之一的同态滤波算法matlab实现(Matlab implementation of homomorphic filtering algorithm, one of the classical image processing methods)

最大熵图像复原方法，效果还好，这是最近努力的结果，不过或许需要适当的改进之后会有更好的结果，分享一下(Maximum entropy image restoration method, the effect Fortunately, this is the result of recent efforts, but may need to be appropriate to improve after the better results, share)

实现图像融合的一些基本代码 实现了梯度金字塔拉普拉斯金字塔算法的图像融合 (matlab This procedure based on the implementation of the FSD pyramid image fusion algorithm, the effect was not bad, you can reference.)

仿射变换来实现旋转 (线性变换)，平移 (向量加)，缩放操作 (线性变换)(Affine transformations are implemented by rotation (linear transformation), translation (vector addition), scaling operation (linear transformation))

说明：  傅立叶变换在图像处理中的重要作用： (A) 图像增强与图像去噪 绝大部分噪音都是图像的高频分量，通过低通滤波器来滤除高频噪声；边缘也是图像的高频分量，可以通过添加高频分量来增强原始图像的边缘； (B) 图像分割与边缘检测 提取图像高频分量 (C) 图像特征提取 形状特征：傅里叶描述子 纹理特征：直接通过傅里叶系数来计算纹理特征 其他特征：将提取的特征值进行傅里叶变换来使特征具有平移、伸缩、旋转不变性 (D ) 图像压缩 可以直接通过傅里叶系数来压缩数据；常用的离散余弦变换是傅立叶变换的实变换； (Fourier transform in the important role of image processing: (A) image enhancement and image denoising Most of all images of high frequency noise by low pass filter to filter out high frequency noise edge is the image of the high frequency components can be enhanced by adding high frequency components to the edge of the original image (B) Image segmentation and edge detection High frequency components extracted image (C) image feature extraction Shape features: Fourier descriptors Texture: directly through the Fourier coefficients to calculate the texture features Other features: The characteristic value extracted by Fourier transform to the characteristics of translation, scaling, rotation invariance (D) image compression Fourier coefficients can be directly compressed data commonly used discrete cosine transform is the Fourier transform of the real transformation 朗读显示对应的拉丁字符的拼音)

根据RBM深度网络（Hinton，2006）进行MINST手写文字的分类器的训练，利用训练得到的权值制作了这个小程序，通过这个程序可以看出训练结果对数据集内的测试样本和训练样本都能进行很好的识别，但是对其他的手写字体识别就没有那么好了。(According to RBM depth network (Hinton, 2006) conducted MINST handwritten text classifier training, using the training to get the right value to produce this small program, this program can be seen through the results of the training data set and the training sample test samples are can be well identified, but on the other handwriting recognition is not so good.)

维纳滤波，图像恢复，通过用matllab编辑的程序实现以上功能-wena (Wiener filtering, image restoration, with matllab editing program to achieve the above functions-wena)

说明：  对多帧图像进行超分辨率处理，得到一副高分辨率的单帧度图像。(Multi-frame image super-resolution processing, get a degree of high-resolution single-frame image.)

基于平稳小波变换的图像融合。低频分量取平均，高频分量取绝对值最大。最后做逆平稳小波变换。(stationary wavelet transfrom.Averaged low frequency components, high frequency components of the maximum absolute value. Finally, do the inverse stationary wavelet transform.)