彩色立体图像匹配算法研究

Matching Algorithms of a Pairs of Color Stereo Images

何友兵[1] 李大海[1] 李良玉[2] 李付明[1] 刘泽晟[1] 王琼华[1]

立体图像匹配对于自由立体重建、三维测量以及立体显示中图像压缩来说是非常重要的.针对区域灰度互相关图像匹配的计算量大、速度慢的情况,运用了一组快速灰度互相关计算公式,提出了用视差限制搜索空间.同时,针对彩色图像的特点,将彩色图像的色彩信息考虑进图像的匹配过程中,将彩色图像的R、G、B三单色图像分别进行灰度相关匹配,定义并采用了颜色分量权重系数来抑制误匹配点.实验表明,该方法有效地减少了计算量,加快了计算速度,提高了图像匹配精度,具有较好的实时性.该方法将应用于三维立体显示中基于视差图的视差进行图像压缩.[著者文摘]

The matching of a pairs of stereo images is very important for auto stereo reconstruction and stereo measurement. In this article, a group of fast grey cross-correlation calculation formula is applied to solve the mass consuming time calculation in process of stereo images matching. And the maximal parallax scope of a pairs of stereo images is used to reduce the search region of matching calculation. At the same time, according to the characteristics of colour images, the colour information is took into account in image matching, and this two color images are split into three monochromatic images （R, G, B）, respectively. And then each monochromatic image is matched respectively. In addition, the colour weight coefficient is adopted to eliminate ambiguity. The experiment results show that the algorithm can reduce the amount of calculation to accelerate the matching calculation speed and the accuracy of image matching is increased. Therefore, the algorithm has well real-time performance and robustness. For purpose, the algorithm will be applied to image compression on based parallax among stereo imagesin three dimensions stereo display.[著者文摘]

对数极坐标图像匹配在目标姿态测量中的应用

Log-polar image matching for moving target attitude detection

张叶[1,2] 曲宏松[1,2] 赵建[1] 王延杰[1]

本文介绍了一种基于极坐标下图像匹配的方法实现帧与帧之间旋转角度的计算,并可以得到目标轮廓保证可靠跟踪,其坐标变换部分不同于传统的对数极坐标变换,该方法并不进行对笛卡尔坐标系下图像的采样,而是完全保留目标精确的边缘信息用以进行边缘形状匹配。匹配过程分为粗匹配和精匹配两步进行,这种方法主要是利用了极坐标下目标轮廓旋转不变性来实现的。本文还提出了一种双目测量目标旋转角度系统,通过该系统可以完成对空间中的目标在任一旋转平面上的旋转角度。实验证明,该方法计算量小、简单实用、跟踪精度满足要求。[著者文摘]

The target attitude extraction is an important task for moving target trackers, especially, the rotated targets recognition and tracking. A technique to calculate rotating angles and to utilize scaling value to ensure credibility tracking with contour of the target is introduced in the paper. In this method, the coordinate transformation is different from the traditional log-polar trasform. The accurate contour is remained completely after the coordinate transformation. The matching process is divided into two steps; one is rough matching and the other is fine matching. And a binocular camera system is proposed in this paper. The experimental results show the tracking accuracy meets the system requirement with a small quantity of calculation.[著者文摘]

基于奇异值分解的图像匹配和目标跟踪研究

Research on Image Matching and Target Tracking Through Singular Value Decomposition

史泽林[1,2] 张志佳[1,2] 黄莎白[1]

在目标跟踪过程中，由于存在目标姿态变化和背景干扰，在跟踪过程中必须要对模板进行必要的修正，应该在获取目标模板后有一个可靠的模板更新策略．本文探讨了奇异值分解及其在图像匹配和目标跟踪中的应用，定义了一种近似奇异值向量并提出了一种基于图像奇异值特征向量数据特点的跟踪策略并作了大量试验，试验结果表明所提出的图像匹配方法和目标跟踪策略的有效性．[著者文摘]

The paper probes into the singular value decomposition and its applications in image matching and target tracking. The singular value has robust performance that is invariant to image disturbance and it makes the proximate singular value vector defined in the paper credible to represent the image as an algebraic feature. The template image should be updated adaptively in respect that the target and background are changing during the process of target tracking. A template-updating strategy is proposed based on the characters of the singular value vector. Experiments are performed on a large test set and the results show that the proposed strategy is practical and efficient in target tracking.

基于遗传算法的导航实时图像匹配算法

Real-time image matching for navigation system based on genetic algorithm

冷雪飞[1] 刘建业[2] 熊智[2]

由于一般图像匹配算法均采用全局搜索法，耗时较大，为满足景象匹配辅助导航系统实时性的要求，提出了一种将遗传算法和加权Hausdorff距离算法相结合的图像匹配算法，利用遗传算法的非遍历搜索机制，迅速收敛到全局近似最优解，提高了匹配搜索的快速性。同时，提出了一种基于特征图像分支点提取的加权Hausdorff距离图像匹配算法，并给出了相应的权值求解公式，利用加权Hausdorff距离作为遗传算法的适应度函数，能够明显减少匹配搜索的计算量，提高匹配结果的精度。仿真分析表明，将遗传算法和加权Hausdorff距离算法相结合的图像匹配算法能够很好地满足景象匹配辅助导航系统的实时性和精度要求。[著者文摘]

In order to meet requirement that scene matching aided navigation system must get aircraft position error real-timely, an image matching algorithm based on genetic algorithm and weighted Hausdorff distance was proposed. In general, the global search in conventional image matching is highly time consuming task. For the non-ergodic search characteristic of genetic algorithm was utilized, the global approximate optimum solution was approached rapidly. Therefore the rapidity of matching search can be improved by our algorithm. Moreover, the weighted Hausdorff distance algorithm based on bifurcations extraction and the corresponding weight formula was proposed. The calculation of matching search can be reduced and the accuracy of matching results can be improved by using the weighted Hausdorff distance as the fitness function of genetic algorithm. Simulation results show that, the proposed image matching algorithm combined with genetic algorithm and weighted Hausdorff distance can satisfy the real-time and accuracy demands of the scene aided navigation system.[著者文摘]

基于差分有序数组的图像匹配快速算法

Fast Image Matching Algorithm based on Differential Ordinal Array

沙莎 刘锦峰

本文提出了一种对模板匹配算法进行改进的快速算法。首先，对模板内所有像素进行排序并差分变换为函数F1（），将模板覆盖下的子图像函数f（x，y）累进求和变换为函数F2（），然后求取F1（）与F2（）乘积的最大值。由于模板存在大量灰度值相同的像素，经排序差分后F1（）中会有很多0和1，乘1和0的运算可以不做，从而消去了模板运算中的大量乘法和加法运算，同时在模板匹配移动过程中利用相邻窗口间的数据相关性，减少重复运算，和传统匹配算法相比，计算复杂度大大降低。[著者文摘]

To improve the traditional algorithm, a fast image matching method is presented, First, sort the template and differential transform to be function F1（）, image which under the template transform to F2（） by recurrence sum f（x,y）, Second, select the max production of F1（） and F2（）. Pixels in template are the same in a large number, after sorting and differencing, there are a lot of 0 and 1, so reduce lots of operation about multiplication and addition in template matching algorithm, at the time, using the relativity of the neighbor window to reduce the repeat operation in the process of template moving, compared with the traditional algorithm, it＇ s make a big reduce in computational complexity.[著者文摘]

一种基于局部投影熵的图像匹配新算法

刘雅轩 苏秀琴 王萍

将图像熵和投影特征的概念结合引入到图像匹配中，定义了图像局部投影熵，并提出一种基于局部投影熵的图像匹配方法．基于局部投影熵的图像匹配方法具有较好的抗几何失真能力．并且提高了抗噪能力和在强光照条件下的匹配能力．结合采用分块、序贯检测及分层搜索等技术，进一步减少了计算量，实验结果表明这是一种简单而行之有效的图像匹配方法．

基于对应像素距离度量的图像匹配跟踪算法

THE ADAPTIVE IMAGE MATCHING TRACKING ALGORITHM BASED ON CORRESPONDING PIXEL DISTANCE MEASUREMENT

江和平 陈洪光 李飚 沈振康

结合最小绝对差度量MAD和Hausdorff距离度量的基本思想，提出了一种新的对应像素距离相似性度量方法的图像匹配算法和自适应模板图像匹配跟踪算法，并对对应像素距离相似性度量方法的图像匹配算法的匹配跟踪性能进行分析．对这2种算法的匹配跟踪性能参数的实验结果进行比较．实验结果表明：对应像素距离相似性度量方法的图像匹配算法具有较强的抗噪、抗畸变能力和稳定性，具有较高的匹配精度和匹配概率．[著者文摘]

The idea of the MAD is combined with the Hausdroff distance measurement in the paper, then a new corresponding pixel distance measurement （CPDM） image matching algorithm and an adaptive template image matching tracking algorithm is presented. The performance of the CPDM image matching algorithm is analyzed. The experiment result of the registration tracking performance with the two algorithms is compared. The results show that the CPDM image matching algorithm has character of the better probability, precision and robustness of registration in the noise, deformation and occlude.[著者文摘]

一种基于图像特征点的图像匹配算法

Image registration on image feature

戚世贵[1] 戚素娟[2]

图像匹配技术被广泛用于人脸识别、全景图像生成等领域。该文利用变比不变特征点（Scale Invariance Feature Transform-SIFT）提取方法提取特征点，并对SIFT方法提取出的特征点用最近邻算法（Nearest Neighbor-NN）进行匹配，在搜索最近邻特征点和次近邻特征点时使用了在K—D树搜索算法基础上进行改进的搜索算法BBF（Best Bin First）算法。实验证明该匹配算法具有匹配精度高，鲁棒性好的特点。[著者文摘]

Image matching technology has been widely used for face recognition, building panorama. This paper uses SIFT （Scale Invariance Feature Transform） as feature extraction method. After that, the paper uses NN （Nearest Neighbor） for feature matching. In the period of searching nearest and second nearest feature point, the paper uses BBF （Best Bin First） as searching method which is modified from K-D tree searching method. This experiment proved that it is of high accuracy and robustness.[著者文摘]

基于临界特征点的图像匹配算法

Image Matching Algorithm Based on Critical Feature Points

刘曙 罗予频 杨士元

基于特征的图像匹配相关算法尽管已经十分普遍并得到广泛应用,但特征的提取容易受噪声影响。该文提出了一种用尺度空间下的临界特征点对图像进行匹配的方法。该方法采用尺度空间下的临界特征点来描述图像的灰度特征,对光照和噪声具有一定的鲁棒性。考虑到不同尺度下特征点对视觉影响的不同,算法用PTD距离对带权重的图像的特征点集进行匹配。由于PTD距离满足三角不等式规则,该算法适合于在大量数据库中快速检索及识别物体。实验证明了该算法的有效性。[著者文摘]

　Algorithms based on image features are very popular and widely used in image matching.However,the feature extraction process is often sensitive to noises.This paper presents an image matching algorithm using critical feature points in space-scale,which represent image gray-level feature.The algorithm is robust to the illumination intensity and noises.For the purpose of comparing distance between weighted feature points,the proportional transportation distance is used.Because PTD obeys the triangle inequality,the algorithm is suitable for efficient object retrieval and recognition in large database.Experiment result confirms the efficiency.[著者文摘]

基于DOG特征点的序列图像匹配算法

Serial Images Matching Algorithm Based on DOG Feature Points

唐永鹤[1] 卢焕章[1] 侯文杰[2]

针对实时匹配的要求，提出一种基于DOG特征点的快速图像匹配算法，用以匹配只存在平移和较小旋转的序列图像。该算法通过求高斯差分算子在尺度空间上的局部极大值和极小值提取特征点，然后根据圆旋转不变特性生成20维的旋转不变特征描述子，并充分利用特征点的区域特征和灰度特征进行匹配，最后根据序列图像对应特征点之间的距离基本保持不变的特性剔除错误的匹配点。实验结果表明该算法快速有效，而且对噪声影响不敏感，具有很强的实用性。[著者文摘]

In order to match in real-time,a fast algorithm based on DOG feature points is presented in this paper, which can be used to find corresponding points between serial images in the case of translation and minor rotation. Feature points are extracted in the process of searching for the local maximum and minimum of difference of Gaussian function in scale space. Then rotation invariant feature descriptors of 20 dimensions are generated based on that the shape of circle kept same when rotated. Subsequently,matching operations are performed by making full use of the location and gray-value of feature points. At last,the false matching points are removed according to the characteristic that the distance of corresponding points maintained invariant between serial images. The algorithm is proved to be fast, robust to noise,effective and practical by the experiment results.[著者文摘]