Description: 针对图像配准中常出现的RST(旋转-比例-平移)变换，推导出了相应的盒距离变换公式。与传统的基于广
义仿射变换的Hausdorff盒距离变换公式相比，缩小了搜索距离空间。在计算Voronoi表面时，根据Hausdorff距
离的计算需要提出比较滑动窗口的区域Voronoi表面，节省了计算Voronoi表面的时间。并且在利用边缘点计算
Hausdorff距离时，剔除琐碎的边缘，仅使用较长的边缘计算。试验结果表明，这些改进方法较大地提高了基于
Hausdorff距离的图像配准的计算速度。
-For image registration often appears RST (rotation- the ratio of- translation) transform, derived from the corresponding transformation formula box. With the traditional affine transform based on generalized distance transform of the Hausdorff formula box compared to the distance to narrow the search space. Voronoi surface in the calculation, according to the calculation of Hausdorff distance sliding window need to compare the regional Voronoi surface Voronoi surface is calculated to save time. Edge points and the calculation in the use of Hausdorff distance, remove the edge of trivial, only the edge of a longer calculation. Test results show that the method greatly improved Hausdorff distance-based image registration of the computing speed. Platform: |
Size: 401408 |
Author:媛媛 |
Hits:
Description: 图像264编解码的搜索的网络上的学习笔记，对于深入研究264编解码以及如何编写修改jm的代码是个很好的参考资料，比较全-264 codec image search on the network to learn the notes, the codec 264 in-depth study and how to prepare amendments to the code jm is a good reference to compare the whole Platform: |
Size: 84992 |
Author: |
Hits:
Description: 思路简要说明：
1、图像二值化
将图片中的各点用0或1表示，1为有效点，0为背景。这里使用的是最大类间方差法
（otsu），在资料中有介绍。
2、去除干扰点
3、分割
将整个的图片分为每个单独的字，在下一步中才能一一识别。
4、与样本库进行对比，寻求最近似匹配
这步是比较核心的地方，由于要识别的图形每次都是随机变化的，我们不能进行完
全匹配识别，所以使用的是‘欧氏距离’来进行最近似匹配，资料中的《自由手写体
数字识别》里面有详细说明。
（样本库文件是按照匹配的特征通过事先编写的程序进行学习得到的）
该识别思路对目前很多验证码有效，识别速度快，正确率基本还可以（在一定程度内
样本量越大正确率越高），不能识别的情况也不少，比如字符粘连，导致程序无法正确
分割，从而识别失败，有朋友介绍神经网络识别方法不错，有空一定要学习下。-A brief description of ideas:
1, image binarization
Pictures of all the points expressed by 0 or 1, 1 for the effective point, 0 for background. Used here is the largest between-class variance
(Otsu), are introduced in the data.
2, removal of interference points
3, partition
The whole picture is divided into each individual character, in the next step in order to identify 11.
4, compared with the sample database, the search might match the recent
This step is to compare the core places, due to the recognition of each of the graphics changes are random, we can not be finished
Identification of the entire match, so the use of the Euclidean distance to be like the recent match, the information in the "free handwritten
Digital Identification, "which is described in detail.
(Sample library features in accordance with the matching pre-prepared through the process of learning to be)
The identification of a lot of ide Platform: |
Size: 874496 |
Author:yangq |
Hits:
Description: 本程序主要四大功能特点:
1、继承了 YUVViewer 可以同时打开多窗口的特点；
2、继承了 Elecard YUV Viewer 可以进行两幅图像对比的特点；
3、增加了宏块信息显示功能；
4、增加了两图对比时同步帧跳转功能：
例如对于错误码流，不同的错误隐藏算法可能造成解码序列长度不同，这时要比较两个序列就很不方便。而该功能正是为了解决这个问题，即以另一序列的显示图像为标准，在当前序列中寻找与其完全相同的图像。 -Four main features of this procedure: 1, inherited YUVViewer can also open multiple windows of the characteristics 2, inherited Elecard YUV Viewer can be characterized by comparing two images 3, an increase of the macro block information display 4, increasing When comparing the two plans simultaneously frame jump function: for example, the error stream for the different error concealment algorithm for decoding the sequence length may result in different time sequences to compare two very inconvenient. And the function is precisely to solve this problem, that is another sequence of display images as the standard sequence search in the current image with exactly the same. Platform: |
Size: 230400 |
Author:沈磊 |
Hits:
Description: 为了提高算法的执行效率, 应用粒子群算法求取图像中任意两点间最短路径来定位目标边界, 并与经典的基于Dijkstra 动态规划图搜索的Liv e Wire 算法进行比较.-In order to improve the efficiency of the algorithm, particle swarm algorithm to strike any of images to locate the shortest path between two object boundaries, and with the classical dynamic programming graph search based on Dijkstra' s algorithm to compare Liv e Wire. Platform: |
Size: 318464 |
Author:莉 |
Hits:
Description: 相似图片搜索.关键技术："感知哈希算法"（Perceptual hash algorithm），它的作用是对每张图片生成一个"指纹"（fingerprint）字符串，然后比较不同图片的指纹。结果越接近，就说明图片越相似。
-Similar image search. Key technologies: perception of the hash algorithm (the Perceptual hash algorithm), it is the role of each image to generate a " fingerprint" (fingerprint) string, and then compare different picture of the fingerprint. The results of the closer picture more similar. Platform: |
Size: 249856 |
Author:wt |
Hits:
Description: ：本文从搜索范围、检索方法、工作原理、结果描述、特点、效果评价等6个方面对5种中外网络图像搜索引擎即Lyco~、
WebSEEK、AmaziIlg Picture Machine、Ixquick、天网中英文FTP进行比较和评析，从而较为详细地介绍了网络图像搜索引擎的原理、优缺点和未来发展趋势。-This article is from the Search and retrieval method works, the results described characteristics, the effect evaluation of six aspects of the five kinds of Chinese and foreign Web image search engine that Lyco WebSEEK compare AmaziIlg Picture Machine, the Ixquick Skynet in English FTPand assessment, a more detailed description of the principles of Web image search engine, strengths and weaknesses, and future trends. Platform: |
Size: 131072 |
Author:K |
Hits:
Description: 把 SIFT 算法应用在牙齿模型图像上，检测牙齿图像的特征点。 方法：首先采用高斯差分算子 DoG 搜索整个图
像的尺度和位置信息，从而确定具有代表性尺度、方向的特征点。基于其稳定性选择关键点，得到一个详细的模型以确定每个候
选点的合适位置和范围。基于局部图像梯度方向信息将方向矢量和关键点对应起来。在选定范围内的每个关键点周边区域测量
局部图像梯度，并采用 KNN 算法进行特征匹配。 结果：通过大量的实验和与其他特征提取方法相比较，该方法能有效地检测牙
齿模型图像的特征，并为牙齿模型三维重建提供有效的参数。-SIFT algorithm is applied to the teeth of the model image, the image feature point detecting teeth. Methods: DoG Gaussian differential operator to search the entire image the scale and location information, to determine a representative scale, the direction of the feature point. Select the key points based on their stability, to get a detailed model to determine the appropriateness of each candidate point location and extent. Information based on local image gradient direction and key points of the direction vectors correspond. Within the selected area around each critical point of measuring the local image gradient, and using KNN algorithm for feature matching. Results: Through a lot of experiments with other feature extraction methods and compare the proposed method can effectively detect tooth model image feature, and to provide an effective three-dimensional reconstruction tooth model parameters. Platform: |
Size: 991232 |
Author:焦婷 |
Hits:
Description: Article is devoted to the use of Haar cascades to compare similar images, object tracking problems between adjacent video frames, there is a match on several photographs, image search on the image and other similar problems. Platform: |
Size: 11350016 |
Author:Andriy |
Hits:
Description: Search image and compare. The source code of the program, showing an example of image retri on a computer and the allocation of duplicate by comparing them - Author unknown.
-Search image and compare. The source code of the program, showing an example of image retri on a computer and the allocation of duplicate by comparing them - Author unknown.
Platform: |
Size: 16384 |
Author:denis76506 |
Hits:
Description: 全搜索算法重构图。像顾名思义，穷尽法就是对搜索范围内的每个位置都进行计算并比较，从原图像块中找到与待匹配块最相似（误差最小）位置的方法。对于一个搜索范围为[-7,7]的区域来说，x，y两个方向全部搜索一遍需要进行 15X15 = 225 次计算。穷尽法的缺点很明显就是计算量太大，速度很慢，另一方面，由于全部范围都进行了计算，其匹配精度是最高的。(Reconstruction graph of full search algorithm. As the name implies, the exhaustive method is to calculate and compare each position within the search range, and find the most similar (least error) position with the block to be matched from the original image block. For an area with a search range of [- 7,7], 15x15 = 225 calculations are required for searching both directions X and Y once. The disadvantage of exhaustive method is that the calculation is too large and the speed is very slow. On the other hand, the matching accuracy is the highest because all the ranges are calculated.) Platform: |
Size: 8902656 |
Author:jiayan |
Hits:
Search - image compare search