gmkl

推荐火灾检测的参考文献，有火焰多个特征提取的算法。(the article is about how to detect fire，using many algorithms to get the features of fire.)

本算法在Matlab2008b 环境下实现。包括main,san 和ssd 三个函数。 这次实现的算法并不是比较两个已经知道的点是否匹配，而是已知一个图形中的特征点，在另外一个图像中找到与其最匹配的点，匹配度用SAD 或者SSD 来度量。 main.m 是程序的入口，包括生成和读入实验数据，调用sad 函数和ssd 函数求匹配点，最后绘 制出最后的结果。sad.m 是用sad 度量方法在另一图中求解匹配点的函数实现，有3 个参数 y=sad(x,image1,image2)。其中x 是代匹配的数据，image1 是图像1 数据，image2 是图像2 数据。 从理论上分析，ssd 比sad 算法要复杂一点，经过测试，在一幅640*480 的图像中寻找10 个匹配点数据SAD 用时25.062519 秒，SSD 用时25.291432 秒。(The algorithm Matlab2008b environment to achieve. Including the main, san, and ssd three functions. The implementation of the algorithm is not the point of comparing two matches already know, but the known feature points in a graphic, an image found in the other match with the most points, matching measured with the SAD or SSD. main.m is the entry, including the generation and reading test data, call the sad ssd function evaluation functions and matching points, and finally draw the final result. sad.m measure is sad figure in another match point in the function implementation to solve, there are three parameters y = sad (x, image1, image2). Where x is the generation of matching data, image1 is the data image 1, image2 is the image 2 data. From the theoretical analysis, ssd little more complicated than the sad algorithm, tested in a 640* 480 images of 10 match points in the search for data using time 25.062519 seconds SAD, SSD with time 25.291432 seconds. )

基于matlab开发的因离焦产生的图像模糊处理，利用盲卷积处理方法，取得很好的效果(blind deconvolation)

说明：  文献 "Deblurring Text Images via L0-Regularized Intensity and Gradient Prior" 的参考代码 用Lp正则化方法做盲复原的代码 demo_text_deblurring 是主函数(refer to "Deblurring Text Images via L0-Regularized Intensity and Gradient Prior" main function: demo_text_deblurring)

区域生长法进行图像分割，通过选取合适的种子点，将感兴趣区域分割出来。(The region growing method is used for image segmentation, and the region of interest is segmented by selecting appropriate seed points.)

说明：  用canny算子来检测图像边缘，VC++(Using canny operator to detect image edge)

经过测试的基于sift程序的图像拼接。以lowe的程序为sift提取核心程序，此外还包括ransac算法，单应矩阵提取，图像融合（包括加权融合以及平均融合）算法，融合效果在testnew中可见。(Image mosaic based on sift program tested. The lowe program sift extract the core program, in addition to including ransac algorithm homography extract image fusion (including fusion weighted average fusion) algorithm, fusion effect is visible in testnew.)

能够用 matlab确定模最大值，利用小波变换去噪，对地震信号进行处理(The maximum modulus can be determined by MATLAB, and the seismic signal is processed by wavelet transform denoising.)

介绍了DAB 接收机中的OFDM 同步算法，重点对OFDM 接收机方案中基于保护间隔的符号定时估计、频 率偏移估计进行了分析，提出了适用于Eureka147DAB 系统的同步算法；并对各种算法进行了计算机仿真，本文介绍的算法满 足DAB 接收机的要求，有很大的应用价值。(Introduced the DAB the OFDM receiver synchronization algorithms, focusing on OFDM receiver scheme based on the protection of the interval estimation, frequency symbol timing Rate migration estimates are analyzed and some appropriate Eureka147DAB system in the synchronization algorithms And kinds of the computer simulation algorithm, this paper introduces the algorithm of the full The requirements of the foot DAB receiver, a large application value. )

输入一个带权无向图的顶点数、边数、各条边信息（两个顶点和权值），建立该图的邻接矩阵结构，输出该邻接矩阵。将上述无向图邻接矩阵转换为邻接表结构，输出该邻接表；根据该邻接表对无向图进行深度优先遍历序列和广度优先遍历序列，并输出遍历结果；用prim算法实现构造该带权无向图的最小生成树，并将该最小生成树的各条边信息输出。(Enter a weighted undirected graph with vertices, edges, each of the side information (two vertices and weights), the establishment of the adjacency matrix structure, the output of the adjacency matrix. The above undirected graph adjacency matrix into adjacency list structure, the output of the adjacency list according to the adjacency table undirected graph depth-first traversal sequence and breadth-first traversal sequence, and outputs through the results using prim algorithm construct the weighted undirected graph minimum spanning tree, and the minimum spanning tree information on respective sides of the output.)