A fast and fully automatic registration approach based on point features for multi-source remote-sensing images

Automatic registration of multi-source remote-sensing images is a difficult task as it must deal with the varying illuminations and resolutions of the images, different perspectives and the local deformations within the images. This paper proposes a fully automatic and fast non-rigid image registration technique that addresses those issues. The proposed technique performs a pre-registration process that coarsely aligns the input image to the reference image by automatically detecting their matching points by using the scale invariant feature transform (SIFT) method and an affine transformation model. Once the coarse registration is completed, it performs a fine-scale registration process based on a piecewise linear transformation technique using feature points that are detected by the Harris corner detector. The registration process firstly finds in succession, tie point pairs between the input and the reference image by detecting Harris corners and applying a cross-matching strategy based on a wavelet pyramid for a fast search speed. Tie point pairs with large errors are pruned by an error-checking step. The input image is then rectified by using triangulated irregular networks (TINs) to deal with irregular local deformations caused by the fluctuation of the terrain. For each triangular facet of the TIN, affine transformations are estimated and applied for rectification. Experiments with Quickbird, SPOT5, SPOT4, TM remote-sensing images of the Hangzhou area in China demonstrate the efficiency and the accuracy of the proposed technique for multi-source remote-sensing image registration.     

Unsupervised deformable image registration network for 3D medical images

Image registration aims to establish an active correspondence between a pair of images. Such correspondence is critical for many significant applications, such as image fusion, tumor growth monitoring, and atlas generation. In this study, we propose an unsupervised deformable image registration network (UDIR-Net) for 3D medical images. The proposed UDIR-Net is designed in an encoder-decoder architecture and directly estimates the complex deformation field between input pairwise images without any supervised information. In particular, we recalibrate the feature slice of each feature map that is propagated between the encoder and the decoder in accordance with the importance of each feature slice and the correlation between feature slices. This method enhances the representational power of feature maps. To achieve efficient and robust training, we design a novel hierarchical loss function that evaluates multiscale similarity loss between registered image pairs. The proposed UDIR-Net is tested on different public magnetic resonance image datasets of the human brain. Experimental results show that UDIR-Net exhibits competitive performance against several state-of-the-art methods.

     

Registering multiple range images based on distance metric of surface-to-surface

1 Introduction In recent years, there has been growing interest in the range sensing techniques for building 3D computer models of real-world objects and scenes without requiring hu-mans to manually produce these models using laborious and error-prone CAD-based approaches. Using range sensors, users are able to capture 3D range images of objects from different viewpoints that may be combined to form the final model of the object or scene[1]. These models then may be used for a variety of app…     

基于点特征的多源遥感影像高精度自动配准方法

提出了一种基于点特征的多源遥感影像高精度自动配准方法。该方法采用了由粗到精的配准策略。首先利用SIFT算子和一次多项式实现影像的粗配准,粗配准后的影像和参考影像将处于同一尺度（像素采样间隔）和参考坐标系下。其次在粗配准后的影像上提取分布均匀的特征点,根据前一步得到的影像间的坐标关系,在参考影像上确定一个较小的搜索范围,使用相关系数匹配出同名点,同时用Baarda数据探测法剔除粗差。最后根据同名点构建三角网对影像进行精配准。实验结果表明:该方法能够实现多源遥感影像的高精度配准。     

基于形变模型由立体序列图象恢复物体的3D形状

结合立体视觉和形变模型提出了一种新的物体3D形状的恢复方法。采用立体视觉方法导出物体表面的3D坐标;利用光流模型估计物体的3D运动,根据此运动移动形变模型,使其对准物体的表面块;由形变模型将由各幅图象得到的离散的3D点融为一起,得到物体的表面形状。实验结果表明该方法能用于形状复杂的物体恢复。     

对数极坐标特征指导的迭代就近点法

在多模态图像,有光照、取向及尺度或纹理变化的图像,以及宽基线图像之间通常存在大的形貌畸变。然而,当前主流推广的双自举迭代就近点法(GDB-ICP)在配准这类图对时存在困难。主要原因是,该方法利用提取的尺度不变泡点(SIFT keypoint)来引导迭代就近点法(ICP),但在大畸变图像上提取的SIFT点是不可靠的。为此,提出了一种用对数极坐标特征点(LPF)来引导迭代就近点的图像配准新方法(LPF-ICP)。实验结果表明,LPF-ICP方法成功地从Rensselaer数据组中的所有22对挑战性图对提取了可靠的LPF种子,并顺利实现了全图配准,而GDB-ICP方法则只完成了其中的19对,从而证实了LPF-ICP方法的有效性。     

结合SIFT和Delaunay三角网的遥感图像配准算法

针对高分辨率遥感图像中提取的特征点数目过大且易存在误匹配点的问题,提出了一种粗配准和精配准相结合的高分辨率遥感图像配准算法.首先对图像降采样处理后,提取大尺度空间下的SIFT特征点,求得仿射变换模型完成图像粗配准;然后对图像进行分块,利用SIFT方法对每幅子块图像提取特征点,并找到对应子块图像之间的匹配点对;之后利用特征点构建Delaunay三角网,计算每对子块图像之间的三角形相似度,构成相似矩阵,从中挑选相似度大的三角形对以构成精确匹配点对;最后利用得到的精确匹配点对实现最终的图像配准.该算法能够减少提取的特征点数且剔除更多的错误匹配点,从而进一步提高精确匹配点率.实验结果表明了算法的有效性.     

基于高通能量变换的多传感器图像配准

针对多源传感器图像的特点,提出了一种结合高通能量变换与角点匹配的图像配准方法。该方法首先利用SUSAN 算法进行角点检测,然后将角点匹配分为粗匹配和精匹配两个阶段实现：粗匹配时,对图像进行高通能量变换,有效减少多传感器图像间的差异,并利用高通能量图上角点邻域间的归一化互相关度量,来建立角点初始对应关系;精匹配时,通过RANSAC算法,进一步筛选出正确的匹配点对,并获取匹配参数。实验证明该方法能快速、准确地配准红外与可见光图像。     

结合纹理信息的深度像匹配

利用图像的相关性在不同深度像对应的纹理图中自动选取特征点对;然后根据Hausdorff距离判断对应点对的有效性,并利用这些对应点对获得相应的深度点对,从而计算出2个深度像之间的初始位置变换关系;最后结合一种改进的带纹理的ICP算法实现深度像的精确配准.实验结果证明:该算法不仅适用于自由曲面的深度像匹配,而且适用于规则对称物体的匹配.     

基于投影图像 SURF 特征提取的三维模型配准

针对传统三维模型配准方法存在对点云初始位置有一定要求、模型配准的精度有 时不高等问题,提出了一种基于三维模型投影图像 SURF 特征提取的三维模型配准方法。首先 通过扫描三维模型数据确定投影图像的范围,判断每个投影图像像素所隶属的模型网格,并求 解从投影图像到纹理图像的映射关系,从而获取二维投影图像;然后对这两幅投影图像分别进 行 SURF 特征点的选取与特征值的计算,并按 SURF 特征值进行特征匹配,再根据投影图像像 素点与三维网格端点的映射关系计算三维特征点对;最后通过匹配的特征点对求取模型变换矩 阵完成三维模型的配准。实验结果表明,该方法在配准时间变化不大的前提下,有效提高了配 准精度,并具有较好的鲁棒性。     

Fourier–Mellin registration of two hyperspectral images

Hyperspectral images contain a great amount of information which can be used to more robustly register such images. In this article, we present a phase correlation method to register two hyperspectral images that takes into account their multiband structure. The proposed method is based on principal component analysis, the multilayer fractional Fourier transform, a combination of log-polar maps, and peak processing. The combination of maps is aimed at highlighting some peaks in the log-polar map using information from different bands. The method is robust and has been successfully tested for any rotation angle with commonly used hyperspectral scenes in remote sensing for scales of up to 7.5× and with pairs of hyperspectral images taken on different dates by the Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) sensor for scales of up to 6.0×.     

基于SIFT与Contourlet变换的高分辨遥感图像配准

针对高分辨遥感图像特征量较多的情况,提出一种基于SIFT与Contourlet变换相结合的图像配准算法。首先将图像进行Contourlet变换分解成低频和高频子带,对高频子带通过设定合适的阈值来提取图像边缘特征点,对低频子带进行SIFT特征点提取。将两者提取到的特征点分别匹配后得到粗匹配点对,利用随机抽样一致性(RANSAC)选择出精匹配点对,实现图像配准。实验表明:在多源遥感图像配准过程中,与基于非采样Contourlet变换(NSCT)和基于SIFT特征提取相比,该算法能够更准确地提取到特征点,具有更高的运算效率以及匹配率。     

基于特征相似性的RGBD点云配准

三维点云数据的配准是计算机视觉领域的重要研究课题,也是三维重建的关键步 骤。针对 RGBD 点云数据的配准问题,提出一种基于特征相似性的初始配准方法。首先需要计 算待配准的 RGBD 点云模型的曲率和颜色特征度(CFD),并对 CFD 进行统计分析,若模型颜色 特征足够丰富优先采用颜色相似性策略,反之尝试曲率相似性策略。通过特征点提取精简点云 模型,利用确定的对应点选择策略选择候选对应点对。在候选对应点对上采用优化样本一致性 算法获得初始配准变换矩阵,实现两片点云的初始配准。针对不同颜色纹理的 RGBD 点云模型, 本文方法可以自适应选择合适的特征点选择策略,实现点云间良好的初始配准。实验结果表明, 对于几何特征不明显的 RGBD 模型,本文方法能够自适应选择颜色相似性策略来较好地完成初 始配准。对于不同类型的模型配准结果较好,算法效率更高。     

Robust point correspondence applied to two- and three-dimensionalimage registration

Accurate and robust correspondence calculations are very important in many medical and biological applications. Often, the correspondence calculation forms part of a rigid registration algorithm, but accurate correspondences are especially important for elastic registration algorithms and for quantifying changes over time. In this paper, a new correspondence calculation algorithm, CSM (correspondence by sensitivity to movement), is described. A robust corresponding point is calculated by determining the sensitivity of a correspondence to movement of the point being matched. If the correspondence is reliable, a perturbation in the position of this point should not result in a large movement of the correspondence. A measure of reliability is also calculated. This correspondence calculation method is independent of the registration transformation and has been incorporated into both a 2D elastic registration algorithm for warping serial sections and a 3D rigid registration algorithm for registering pre and postoperative facial range scans. These applications use different methods for calculating the registration transformation and accurate rigid and elastic alignment of images has been achieved with the CSM method. It is expected that this method will be applicable to many different applications and that good results would be achieved if it were to be inserted into other methods for calculating a registration transformation from correspondences     

基于多核并行和动态阈值的点云配准算法

针对点云配准中存在错误匹配点对、精度不高等问题,提出一种基于多核并行和动态阈值的点云配准算法。该算法采用改进的SAC-IA算法进行点云粗配准,利用OpenMP实现点云查询点的法向量、FPFH等特征的并行加速提取以及对应点对的并行查找,从而使整个配准算法的速度得到保持甚至提升。在点云精配准阶段,使用改进的ICP算法进行精配准,改进点着眼于错误对应点对的剔除及其阈值的动态确定,即以配准点重心作为参照点,按照动态阈值,使用点对距离约束剔除错误对应点对。实验结果表明,本文算法在提升配准精度的情况下,配准速度也得到了提升。     

遥感影像空间分治快速匹配

目的 图像匹配是遥感图像镶嵌拼接的重要环节,图像匹配技术通常采用两步法,首先利用高维描述子的最近和次近距离比建立初始匹配,然后通过迭代拟合几何模型消除错误匹配。尽管外点过滤算法大幅提高了时间效率,但其采用传统的两步法,构建初始匹配的方法仍然非常耗时,导致整个遥感图像拼接的速度提升仍然有限。为了提高遥感图像匹配的效率,本文提出了一种基于空间分治思想的快速匹配方法。方法 首先,通过提取图像的大尺度特征生成少量的初始匹配,并基于初始匹配在两幅图像之间构建成对的分治空间中心点;然后,基于范围树搜索分治空间中心点一定范围内的相邻特征点,构造成对分治空间点集;最后,在各个分治空间点集内分别进行遥感图像特征的匹配。结果 通过大量不同图像尺寸和相对旋转的遥感图像的实验表明,与传统的和其他先进方法相比,本文方法在保证较高精度的同时将匹配时间缩短到1/1001/10。结论 利用初始种子匹配构建分治匹配中心以将图像匹配分解在多个子区间进行的方法有助于提高遥感影像匹配的效率,该算法良好的时间性能对实时遥感应用具有实际价值。     

基于3D-Harris 与FPFH 改进的3D-NDT 配准算法

针对传统点云配准三维正态分布变换(3D-NDT)、迭代最近点(ICP)算法在未给定初 始配准估计的情况下配准效果不佳、配准时间长、误差较大的缺陷,提出了精准且相对高效的 点云匹配算法。首先,运用3D-Harris 算法识别每一幅点云的关键点,并以此为基本点建立局 部参考框架,计算快速点特征直方图(FPFH)描述子;之后,使用最小中值法(LMeds)中的对应 估计算法排除不准确的点对应关系,得到含有对应三维特征关系的特征点对。计算粗配准所需 的变换矩阵,完成初步匹配。随后,根据3D-NDT 算法将点云数据空间体素化,运用概率分布 函数完成最终的点云进行精确地匹配。使用改进配准将3 组分别从网络下载的较少噪声、大规 模与Kinect V2.0 采集的较多噪声、大规模的2 组重叠度不同的点云数据匹配到同一个空间参考 框架中,并通过精度分析对比经典3D-NDT,ICP 等算法。实验结果证明,该算法在迭代次数 较低时,可使室内场景点云数据完成精度较高的配准且受噪声影响较小,但如何将算法的复杂 度适当降低,缩短配准时间需要更进一步的研究。     

多视点距离图像的对准算法

提出一种多视点距离图像的对准算法.该算法将有拒绝的随机抽样和迭代最近点 (ICP: Iterative Closest Point)算法结合起来,采用粗、精对准时不同的评价函数,利用最小二 乘进行多视点之间运动参数的估计.为了快速进行3D点到物体表面的最近距离和最近点的 计算,采用了物体表面的八叉树样条表示.实验结果表明,该对准算法收敛速度较快,抗噪声 能力较强,并且有较高的对准精度.     

Simultaneous nonrigid registration of multiple point sets and atlas construction

Group-wise registration of a set of shapes represented by unlabeled point-sets is a challenging problem since, usually this involves solving for point correspondence in a nonrigid motion setting. In this paper, we propose a novel and robust algorithm that is capable of simultaneously computing the mean shape represented by a probability density function from multiple unlabeled point-sets and registering them non-rigidly to this emerging mean shape. This algorithm avoids the correspondence problem by minimizing the Jensen-Shannon (JS) divergence between the point sets. We motivate the use of the JS divergence by pointing out its close relationship to hypothesis testing. We derive the analytic gradient of the cost function in order to efficiently achieve the optimal solution. JS-divergence is symmetric with no bias toward any of the given shapes to be registered and whose mean is being sought. A by product of the registration process is a probabilistic atlas defined as the convex combination of the probability densities of the input point sets being aligned. Our algorithm can be especially useful for creating atlases of various shapes present in images as well as for simultaneously (rigidly or non-rigidly) registering 3D range data sets without having to establish any correspondence. We present experimental results on real and synthetic data.     

Automated registration of multi-view point clouds using sphere targets

This paper addresses the problem of automated registration of multi-view point clouds generated by a 3D scanner using sphere targets. First, sphere targets are detected from each point cloud. The centroids of the detected targets in each point cloud are then used for rough registration. Congruent triangles are computed from the centroids for the correspondence among them, with which a rigid body transformation is obtained to bring the two point clouds together as closely as possible. After the initial registration, the two point clouds are further registered by refining the position and orientation of the point clouds using the underlying geometric shapes of the targets. These registration steps are integrated into one system that allows two input point clouds automatically registered with no user intervention. Real examples are used to demonstrate the performance of the point cloud registration.