cluster

数字水印的图像处理方法，很好，做图像水印的同行可以参考(Digital watermark image processing methods, well, do their counterparts in the image watermark can refer to the)

在图像去卷积算法引入先验模型的应用,主要是建立马尔可夫随机场先验模型.很大程度上改进了去卷积的效果,更多的保持了图像的边缘特征.(In image deconvolution algorithm Application of the introduction of a priori model, mainly the establishment of Markov random field a priori model. A large extent, improved the effectiveness of deconvolution, more to maintain the image of the edge feature.)

说明：  图像识别预处理，用主要用于汽车车牌号识别(Image recognition for license plate recognition)

mser算法（Maximally Stable Extremal Regions）代码，非文章作者写的，但是还不错，matlab实现。(mser algorithm (Maximally Stable Extremal Regions) code, non-article the authors write, but also good, matlab realize.)

PIV实验结果后处理,速度场，涡量场，应力场(The postprocessing of PIV observations)

三维点云包围盒精简算法，内包含点云数据与处理结果，直观可用，效果良好。(Three dimensional point cloud bounding box reduction algorithm, which contains point cloud data and processing results, is intuitive and available, and the effect is good.)

基于CFA（颜色滤波阵列）插值检测的图像篡改检测程序(Image tamper detection based on CFA (color filter array) interpolation detecting.)

视频格式转换 这里主要是mpeg格式 转化为avi，yuv等多种格式(Video format conversion here is mainly converted to avi mpeg format)

说明：  小波变换图像增强，经过matlab6.0测试，放心可用(Wavelet image enhancement, after matlab6.0 testing, confidence can be)

孔径雷达图像固有的相干斑噪声严重降低了图像的可解译程度，影响了后续目标检测、分类和识别 等应用。因此，SAR图像的相干斑抑制问题一直是SAR图像应用的重要课题之一。一个理想的去斑算法应该在平 滑的同时保持图像的边缘等细节不受损失，目前存在各种各样的算法，但没有一种方法能够完美的满足这一要求。 本文对SAR图像的相干斑抑制问题进行了全面系统的研究，分析了相干斑的形成原因，总结了目前存在的相干斑 抑制算法的主要思路，介绍了具有代表性的算法，并对各种算法的性能进行了定性分析和比较，给出了去斑算法定 量评估方法(The inherent speckle noise of the aperture radar image seriously reduces the degree of interpretation of the image, and affects the detection, classification and recognition of the subsequent targets. And other applications. Therefore, the problem of speckle suppression in SAR images has been one of the most important topics in the application of SAR images. An ideal speckle algorithm should be flat At the same time, the details such as edges of images are not lost. There are various algorithms at present, but none of them can satisfy this requirement perfectly.)