fre-analysis

对图像处理后，将图像中的目标边缘提取出来，然后添加到原图像上(After image processing, the image of the target edge extracted, and then added to the original image)

本源码是关于图像盲去模糊的算法，去模糊的几种方法中，盲去模糊目前研究比较热门(This source is about blind image deblurring , several defuzzification methods, blind deblurring research more popular)

将matlab处理得到的三维矩阵转化为envi软件可读取的数据格式raw(change the 3-D data from matlab to the raw form ENVI can read)

idl 可视化分析与应用（随书程序），详细的程序与应用。(idl visualization analysis and application (with the book procedures), the detailed procedures and applications.)

本人自己实现的单纯meanshift跟踪算法.按原理编译,适合初学者参考(I own a simple realization of the tracking algorithm meanshift. In accordance with the principle of the compiler, suitable for beginners reference)

说明：  数字图像处理：阈值分割及形态学处理，本程序出色的完成了图形提取，内容包括程序，讲解和运行结果(Digital image processing: segmentation and morphology threshold treatment, the procedure was completed excellent graphics extraction, including procedures and operations on the results of)

这是一个口形特征点粗定位程序，首先对唇部图象进行边缘检测，然后利用几何特征进行定位。(This is a lip feature points coarse positioning procedures, first of all on the lip image edge detection, and then positioning the use of geometrical features.)

压缩感知CS——采用小波变换进行稀疏表示，高斯随机矩阵为观测矩阵，重构算法为SL0算法，对256*256的lena图处理，比较原图和SL0算法在不同采样比例(0.74、0.5、0.3)下的重构效果，并各运行50次，比较算法性能PSNR和每次的运行时间 (Compressed sensing CS- using wavelet transform as sparse representation, Gaussian random matrix as the observation matrix and SL0 algorithm as the reconstruction algorithm. Compare lena figure and the reconstruction results using SL0 algorithm at different sampling ratio (0.74,0.50.3),then each runs 50 times, compare the performance of PSNR and each running time)

说明：  语义分割，训练样本，模型训练，成果展示，深度卷积神经网络(Semantic segmentation, training samples, model training, achievement display, deep convolution neural network)

使用frequency-tuned saliency算法计算图像的saliency值，基于此saliency值将图像中的显著区域提取出来。(Using frequency-tuned saliency saliency value of image algorithm, based on the saliency value of the significant image area extracted.)