Gradient-vector-diffusion-denoising

压缩感知中利用增广拉格朗日方程解最小稀疏正则化的恢复算法(DAL solves the dual problem of (1) via the augmented Lagrangian method (see Bertsekas 82). It uses the analytic expression (and its derivatives) of the following soft-thresholding operation， which can be computed for L1 and grouped L1 (and many other) sparsity inducing regularizers. If you are interested in our algorithm please find more details in our technical report or in my talk at Optimization for Machine Learning Workshop (NIPS 2009).)

说明：  对图形进行二维傅里叶变化后的频谱进行中心化(Centralization of the spectrum after two-dimensional Fourier transform)

图形密码源码程序，例程结合易语言特殊功能支持库，调用API函数判断图形图像的点击顺序。(Graphics password source programs, routines combined with special features easy language support library , call the API function to determine whether a graphic image click sequence.)

本文主要介绍遗传算法的基本理论，叙述遗传算法在图像增强的的主要应用，即将原始图像变得更加清晰，特征变得更加明显。 现今关于图像增强的算法有很多，而这些算法大多是基于退化函数或者点扩展函数的知识进行图像处理的。当图像出现模糊或噪声影响大时，设计出的图像清晰化的效果肯定不够理想，因此有必要对图像进行增强处理。于是，可利用到遗传算法这种成熟稳定的仿生物进化的全局寻优算法，进行图像增强，由于遗传算法控制参数少、自适应度高，将选择该方法对图像退化分别进行不同的清晰化处理. 这样既增强了图像的对比度,又克服了传统直方图均衡化处理所造成的灰度级损失等缺点。 通过实验去表明，遗传算法从全局寻优的角度出发，能够较为精确地估计退化系统函数。和传统的线性增强、直方图均衡进行比较。实验结果表明该方法能改善原图像视觉效果，便于之后的图像分析。 (This paper mainly introduces the basic theory of genetic algorithm, genetic algorithm (ga) in the main application of image enhancement, the original image is more clear, characteristic becomes more obvious. Today about the image enhancement algorithm are many, and most of these algorithms is based on the degradation function or the knowledge of the point spread function of image processing. When the image appears fuzzy or noise influence, designed the image clarity of effect must not ideal, therefore it is necessary to enhance the image. Then, genetic algorithm is available to the mature and stable evolution of the global optimization algorithm, for image enhancement, due to less genetic algorithm control parameters, since the high fitness, will choose the method of image degradation, respectively, to manipulate the different motivation. It can enhance the image contrast, and overcome the traditional histogram equalization processing grayscale loss caused by the faults. Throug)

说明：  基于小波变换的图像融合，可实现两幅图像的融合，融合规则可自定义(Image fusion based on wavelet transform can realize the fusion of two images, and the fusion rules can be customized)

说明：  将图像进行各个角度的旋转，以及对两幅或多幅图像进行叠加(Each rotation angle of the image, and the two or more images are superimposed)

使用说明 使用时打开此例题目录下pic中的图片，然后依次单击按钮“转”、“1”、“2”、“3”、“4”和“5”，就可以实现精确的车牌定位。 具体步骤 1.24位真彩色->256色灰度图。 2.预处理:中值滤波。 3.二值化:用一个初始阈值T对图像A进行二值化得到二值化图像B。 初始阈值T的确定方法是:选择阈值T=Gmax-(Gmax-Gmin)/3,Gmax和Gmin分别是最高、最低灰度值。 该阈值对不同牌照有一定的适应性,能够保证背景基本被置为0,以突出牌照区域。 4.削弱背景干扰。对图像B做简单的相邻像素灰度值相减,得到新的图像G,即Gi,j=|Pi,j-Pi,j-1|i=0,1,…,439 j=0,1,…,639Gi,0=Pi,0,左边缘直接赋值,不会影响整体效果。 5.用自定义模板进行中值滤波 区域灰度基本被赋值为0。考虑到文字是由许多短竖线组成,而背景噪声有一大部分是孤立噪声,用模板(1,1,1,1,1)T对G进行中值滤波,能够得到除掉了大部分干扰的图像C。 6.牌照搜索:利用水平投影法检测车牌水平位置,利用垂直投影法检测车牌垂直位置。 7.区域裁剪,截取车牌图像。()

绘制电脑中原始人声信号的时域、频域图，并在其上加入高斯白噪声，之后设计巴特沃兹滤波器将其滤除。(The draw primitive people in the computer sound signal time domain, frequency domain, and join in on the white Gaussian noise, after design Butterworth filter to filter out.)

一种SAR图像去噪算法，lee，kuan，frost(SAR image denoising，lee，kuan，frost)

图像融合中的平均梯度、相关系数、信息熵、交叉熵、联合熵、均方误差、互信息、信噪比、峰值信噪比、均方根误差、空间频率、标准差、均值、扭曲程度、偏差指数等等(Average gradient, correlation coefficient, information entropy, cross-entropy, joint entropy, mean square error, mutual information, signal-to-noise ratio, peak signal-to-noise ratio, root mean square error, spatial frequency, standard deviation, mean, distortion degree, deviation index, etc. in image fusion)