基于分块和仿射不变性的SIFT图像匹配算法

针对传统SIFT算法在匹配时出现实时性差、匹配量低以及RANSANC算法在剔除SIFT误匹配对时误匹配率高的问题,提出一种基于距离相对性的分块匹配算法和基于仿射不变性的误匹配对剔除算法。首先利用传统SIFT算法提取图像中的特征点;然后采用基于距离相对性的分块匹配算法进行特征匹配得到初始匹配对;由于初始匹配对中存在误匹配,接下来运用基于仿射不变性的误匹配对剔除算法来剔除误匹配对;最后,在不同图像变换下进行仿真实验。实验结果表明,算法在保持SIFT算法鲁棒性的基础上,能够得到更多匹配对,正确匹配率提高了10%左右,并且实时性也得到很大改善。     

基于匹配质量提纯的改进D-Nets算法

针对基于特征的图像配准在较大仿射变形以及存在相似目标情况下适应性不佳的问题,为减少算法的时间开销,提出一种基于匹配质量提纯的改进描述网（D-Nets）算法。首先,通过FAST算法检测特征点,并根据Harris角点响应函数以及网格划分相结合的方式进行筛选;然后,在计算直线描述子的基础上构建哈希表和投票表决,从而得到粗匹配对;最后,采用基于匹配质量的提纯方法剔除误匹配。针对牛津大学Mikolajczyk标准图像数据集进行了实验,结果表明：提出的改进D-Nets算法在尺度、视差和光照变化较大的情况下平均配准精度为92.2%,平均时间开销为2.48 s。与尺度不变特征变换（SIFT）、仿射-尺度不变特征变换（Affine-SIFT）、原始D-Nets等算法相比,提出的改进算法与原始算法的配准精度基本相当,但速度最高可提升80倍,并具有最佳鲁棒性,显著优于SIFT、ASIFT算法,非常适于图像配准应用。     

特征点聚类高精度视差图像拼接

目的 针对不同视点下具有视差的待拼接图像中,特征点筛选存在漏检率高和配准精度低的问题,提出了一种基于特征点平面相似性聚类的图像拼接算法。方法 根据相同平面特征点符合同一变换的特点,计算特征点间的相似性度量,利用凝聚层次聚类把特征点划分为不同平面,筛选误匹配点。将图像划分为相等大小的网格,利用特征点与网格平面信息计算每个特征点的权重,通过带权重线性变换计算网格的局部单应变换矩阵。最后利用多频率融合方法融合配准图像。结果 在20个不同场景图像数据上进行特征点筛选比较实验,随机抽样一致性（random sample consensus, RANSAC）算法的平均误筛选个数为30,平均误匹配个数为8,而本文方法的平均误筛选个数为3,平均误匹配个数为2。对20个不同场景的多视角图像,本文方法与AutoStitch（automatic stitching）、APAP（as projective as possible）和AANAP（adaptive as-natural-as-possible）等3种算法进行了图像拼接比较实验,本文算法相比性能第2的算法,峰值信噪比（peak signal to noise ratio,PSNR）平均提高了8.7%,结构相似性（structural similarity,SSIM）平均提高了9.6%。结论 由本文提出的基于特征点平面相似性聚类的图像拼接算法处理后的图像保留了更多的特征点,因此提高了配准精度,能够取得更好的拼接效果。     

面向像对直线特征匹配的线特征矫正与提纯方法

目的 像对直线特征匹配是计算机视觉的重要研究内容,现有这类匹配方法均存在不同程度的误匹配问题。导致此问题的主要因素包括直线检测结果没有位于图像的真正边缘处、缺乏匹配线对的一致性校验。为此本文提出一种面向像对直线特征匹配的线特征矫正与提纯方法。方法 首先提取像对的边缘特征获得二值化边缘图,通过边缘梯度图及梯度矢量图（GVF）建立梯度引力图。其次,采用直线检测方法提取像对的直线特征,并通过梯度引力图矫正直线位置。最后,采用点特征匹配结果计算像对极线,并结合直线匹配结果确定最后的局部校验特征区域,通过随机抽样一致小邻域范围内特征相似性校验直线匹配结果,从而剔除误匹配直线。结果 对一组宽基线像对进行匹配实验,与直接采用直线匹配算法获得的匹配结果相比,矫正后的匹配结果剔除了大部分误匹配线对,将匹配准确率从50%提高到84%,继续提纯该匹配结果获得了100%的匹配准确率。在另一组宽基线像对的匹配实验中,经本文方法处理后的匹配准确率提高近30%。与前两组实验相比,第3组实验的像对摄影姿态变化不大,仅在尺度上有所区别,经本文方法处理后配准率从92%提高到100%。结论 采用本文方法可以大幅提高像对直线特征匹配的准确率,同时该方法可以很容易对其他直线匹配结果进行校正与提纯,具备较高的实用性。     

一种体视显微图像误匹配消除方法

针对尺度不变特征变换( SIFT)匹配中存在的误匹配问题和立体图像特点,提出一种误匹配消除方法。对体视显微图像进行SIFT特征匹配初步得到匹配对,结合体视显微镜标定参数,计算三维点云坐标。将三维点云分别投影到左、右图像中得到新的匹配对,新投影点的图像坐标分别与原来匹配点的图像坐标相减,生成投影向量集。通过左、右2个投影向量集幅值和方向的异常值剔除,实现误匹配消除。实验结果表明,实验图像的误匹配消除率达到100%,同时不消除正确匹配点,提高了匹配精度。     

基于典型相关分析方法的尺度不变特征变换误匹配剔除

针对尺度不变特征变换(SIFT)描述子仅利用特征点的局部邻域灰度信息而对图像内具有相似灰度分布的特征点易产生误匹配的问题,提出一种基于典型相关分析(CCA)的SIFT误匹配剔除方法.该方法首先利用SIFT算法进行匹配,得到初始匹配对; 然后根据典型相关成分的线性关系拟合直线,利用点到直线的距离剔除大部分误匹配点对; 对剩余的匹配点对,逐一分析其对典型相关成分的共线性的影响,剔除影响程度大的特征点对.实验结果表明,该方法能够在剔除误匹配的同时保留更多的正确匹配,提高了图像配准的精度.     

基于偏最小二乘的SIFT误匹配校正方法

针对尺度不变特征变换（SIFT）描述子仅利用特征点的局部邻域信息而对图像内具有相似结构的特征点易产生误匹配的现象,提出一种基于偏最小二乘的SIFT误匹配校正方法。该方法首先利用SIFT算法进行匹配,得到初始匹配对,然后利用偏最小二乘方法对匹配后初始匹配点的空间分布信息进行重新描述,并通过定义影响函数,剔除影响程度大的特征点对,最后得到精确匹配点对,对图像进行配准。实验结果表明,该方法能够有效地剔除误匹配点,提高图像配准的精度。     

融合特征描述符约束的3维等距模型对应关系计算

目的 为了更准确地构建3维等距模型之间的对应关系,本文提出了一种基于热核签名与波核签名的融合特征描述符计算3维等距模型对应关系的方法。方法 首先计算3维模型Laplace算子获得模型的特征向量和特征值;然后将所得到特征值和特征向量作为基参数分别计算源模型与目标模型的热核签名和波核签名,并将热核签名与波核签名融合为一个新的特征描述符。融合特征描述符作为模型上随机均匀采样点的约束,通过最小值匹配算法得到源模型和目标模型之间的对应关系。结果 实验结果表明,利用融合特征描述符约束进行计算得到的对应关系正确匹配率比热核签名约束计算得到的对应关系匹配率平均提高19.429%,比波核签名约束计算得到的对应关系匹配率平均提高4.857%。结论 本文提出的融合特征描述符适用于计算3维等距模型或近似等距的3维模型之间的对应关系,与单一使用热核签名或波核签名特征描述符相比,可以得到更加准确的对应关系。     

使用霍夫变换的3维点云拼合算法

目的 直接基于点云数据本身的拼合算法对点云模型的位置和重叠度有着较高的要求。为了克服这种缺陷,提出一种针对散乱点云的分步拼合算法。方法 不同于大多数已有的基于曲率信息的拼合算法,本文算法包含了一个序贯式的匹配点对筛选过程和一个基于霍夫变换的坐标变换参数估计过程。在筛选过程中,首先利用曲率相似度确定点云数据之间的初始匹配关系,然后利用刚体不变量特征邻域标识相似度以及持续特征直方图相似度对初始匹配点对进行连续两次筛选以便得到更为精确的匹配点对集。在参数估计阶段,通过对匹配点对的旋转矩阵和平移矢量的参数化处理,利用霍夫变换消除错误匹配点对对坐标变换参数估计的影响,从而得到更加准确的坐标变换参数,实现点云的3维拼合。结果 利用本文算法对两片部分重叠的点云数据进行了拼接实验。实验结果表明,本文算法能很好地实现对部分重叠点云的拼合。由于霍夫变换的引入,本文算法相较于经典的Ransac算法具有更高的正确率、稳定性以及抗噪性,在运行速度上也具有一定的优越性。结论 本文算法不仅能适用于任何具有任意初始相对位置的部分重叠点云的拼接,而且可以取得很高的拼合精度和很好的噪声鲁棒性。     

融合图像感知哈希技术的运动目标跟踪

目的 为解决运动目标跟踪时因遮挡、尺度变换等产生的目标丢失以及传统匹配跟踪算法计算复杂度高等问题,提出一种融合图像感知哈希技术的运动目标跟踪算法.方法 本文算法利用感知哈希技术提取目标摘要进行模板图像识别匹配,采用匹配跟踪策略和搜索跟踪策略相配合来准确跟踪目标,并构建模板评价函数和模板更新准则实现目标模板的自适应更新,保证其在目标发生遮挡和尺度变换情况下的适应性.结果 该算法与基于NCC(normalized cross correlation)的模板匹配跟踪算法、Mean-shift跟踪算法以及压缩跟踪算法相比,在目标尺度变换和物体遮挡时,跟踪的连续性和稳定性更好,且具有较低的计算复杂度,能分别降低跟踪系统约6.2%、 6.3%、 9.3%的计算时间.结论 本文算法能有效实现视频场景中目标发生遮挡及尺度变换情况下的跟踪,跟踪的连续性和稳定性良好,且算法具有较低的计算复杂度,有利于实时性跟踪系统的构建.     

Hough Pyramid Matching: Speeded-Up Geometry Re-ranking for Large Scale Image Retrieval

Exploiting local feature shape has made geometry indexing possible, but at a high cost of index space, while a sequential spatial verification and re-ranking stage is still indispensable for large scale image retrieval. In this work we investigate an accelerated approach for the latter problem. We develop a simple spatial matching model inspired by Hough voting in the transformation space, where votes arise from single feature correspondences. Using a histogram pyramid, we effectively compute pair-wise affinities of correspondences without ever enumerating all pairs. Our Hough pyramid matching algorithm is linear in the number of correspondences and allows for multiple matching surfaces or non-rigid objects under one-to-one mapping. We achieve re-ranking one order of magnitude more images at the same query time with superior performance compared to state of the art methods, while requiring the same index space. We show that soft assignment is compatible with this matching scheme, preserving one-to-one mapping and further increasing performance.     

A General Method for Geometric Feature Matching and Model Extraction

Popular algorithms for feature matching and model extraction fall into two broad categories: generate-and-test and Hough transform variations. However, both methods suffer from problems in practical implementations. Generate-and-test methods are sensitive to noise in the data. They often fail when the generated model fit is poor due to error in the data used to generate the model position. Hough transform variations are less sensitive to noise, but implementations for complex problems suffer from large time and space requirements and from the detection of false positives. This paper describes a general method for solving problems where a model is extracted from, or fit to, data that draws benefits from both generate-and-test methods and those based on the Hough transform, yielding a method superior to both. An important component of the method is the subdivision of the problem into many subproblems. This allows efficient generate-and-test techniques to be used, including the use of randomization to limit the number of subproblems that must be examined. Each subproblem is solved using pose space analysis techniques similar to the Hough transform, which lowers the sensitivity of the method to noise. This strategy is easy to implement and results in practical algorithms that are efficient and robust. We describe case studies of the application of this method to object recognition, geometric primitive extraction, robust regression, and motion segmentation.     

Wrong Matching Points Elimination after Scale Invariant Feature Transform and Its Application to Image Matching

When images are rotated and the scale varies or there are similar objects in the images, wrong matching points appear easily in the scale invariant feature transform (SIFT). To address the problem, this paper proposes a SIFT wrong matching points elimination algorithm. The voting mechanism of Generalized Hough Transform (GHT) is introduced to find the rotation and scaling of the image and locate where the template image appears in the scene in order to completely reject unmatched points. Through a discovery that the neighborhood diameter ratio and direction angle difference of correct matching pairs have a quantitative relationship with the image’s rotation and scaling information, we further remove the mismatching points accurately. In order to improve image matching efficiency, a method for finding the optimal scaling level is proposed. A scaling multiple is obtained through training of sample images and applied to all images to be matched. The experimental results demonstrate that the proposed algorithm can eliminate wrong matching points more effectively than the other three commonly used methods. The image matching tests have been conducted on images from the Inria BelgaLogos database. Performance evaluation results show that the proposed method has a higher correct matching rate and higher matching efficiency.     

基于模糊随机广义霍夫变换的目标匹配

为了提高工业检测中图像匹配精度和速度,提出了一种用于二维目标匹配的新算法--模糊随机广义霍夫变换(FRGHT)。此算法结合了模糊推理系统(FIS)和随机广义霍夫变换(RGHT)。模糊推理系统引入模糊集合概念,计算待配准图像中边缘点对配准参数的投票,从而可以抑制噪声,解决扭曲问题,提高了匹配精度;随机抽取待配准图像中边缘点进行投票,实现了多对一的映射,从而减少了内存需求,提高计算速度。实验表明,该方法计算速度快,匹配精度高,不受噪声污染、扭曲、遮挡、混乱等情况的影响。     

Recovering 3D motion of multiple objects using adaptive Houghtransform

We present a method to determine 3D motion and structure of multiple objects from two perspective views, using adaptive Hough transform. In our method, segmentation is determined based on a 3D rigidity constraint. Instead of searching candidate solutions over the entire five-dimensional translation and rotation parameter space, we only examine the two-dimensional translation space. We divide the input image into overlapping patches, and, for each sample of the translation space, we compute the rotation parameters of patches using least-squares fit. Every patch votes for a sample in the five-dimensional parameter space. For a patch containing multiple motions, we use a redescending M-estimator to compute rotation parameters of a dominant motion within the patch. To reduce computational and storage burdens of standard multidimensional Hough transform, we use adaptive Hough transform to iteratively refine the relevant parameter space in a “coarse-to-fine” fashion. Our method can robustly recover 3D motion parameters, reject outliers of the flow estimates, and deal with multiple moving objects present in the scene. Applications of the proposed method to both synthetic and real image sequences are demonstrated with promising results     

Generalized Hough Transform Using Regions with Homogeneous Color

A novel generalized Hough transform algorithm which makes use of the color similarity between homogeneous segments as the voting criterion is proposed in this paper. The input of the algorithm is some regions with homogeneous color. These regions are obtained by first pre-segmenting the image using the morphological watershed algorithm and then refining the resultant outputs by a region merging algorithm. Region pairs belonging to the object are selected to generate entries of the reference table for the Hough transform. Every R-table entry stores a relative color between the selected region pairs. This is done in order to compute the color similarity and in turn generate votes during the voting process and some relevant information to recover the transformation parameters of the object. Based on the experimental results, our algorithm is robust to change of illumination, occlusion and distortion of the segmentation output. It recognizes objects which were translated, rotated, scaled and even located in a complex environment.     

Demisting the Hough Transform for 3D Shape Recognition and Registration

In applying the Hough transform to the problem of 3D shape recognition and registration, we develop two new and powerful improvements to this popular inference method. The first, intrinsic Hough, solves the problem of exponential memory requirements of the standard Hough transform by exploiting the sparsity of the Hough space. The second, minimum-entropy Hough, explains away incorrect votes, substantially reducing the number of modes in the posterior distribution of class and pose, and improving precision. Our experiments demonstrate that these contributions make the Hough transform not only tractable but also highly accurate for our example application. Both contributions can be applied to other tasks that already use the standard Hough transform.     

基于改进Hough森林的对象检测方法

基于Hough森林的对象检测是隐式形状模型（ISM）的改进,它借助随机森林完成广义Hough变换。为了进一步提高其检测效果,充分利用训练图像中对象位置是已知的知识,改进了经典的偏移量不确定性度量方法,并优化随机森林的投票,使在Hough空间中真正对象的位置获得更多投票和更高的投票值。实验验证了该方法相比于经典的方法,具有更准确的对象检测效果。     

Hough forests for object detection, tracking, and action recognition

Abstract—The paper introduces Hough forests, which are random forests adapted to perform a generalized Hough transform in an efficient way. Compared to previous Hough-based systems such as implicit shape models, Hough forests improve the performance of the generalized Hough transform for object detection on a categorical level. At the same time, their flexibility permits extensions of the Hough transform to new domains such as object tracking and action recognition. Hough forests can be regarded as task-adapted codebooks of local appearance that allow fast supervised training and fast matching at test time. They achieve high detection accuracy since the entries of such codebooks are optimized to cast Hough votes with small variance and since their efficiency permits dense sampling of local image patches or video cuboids during detection. The efficacy of Hough forests for a set of computer vision tasks is validated through experiments on a large set of publicly available benchmark data sets and comparisons with the state-of-the-art.