Stereo reconstruction from multiperspective panoramas

A new approach to computing a panoramic (360 degrees) depth map is presented in this paper. Our approach uses a large collection of images taken by a camera whose motion has been constrained to planar concentric circles. We resample regular perspective images to produce a set of multiperspective panoramas and then compute depth maps directly from these resampled panoramas. Our panoramas sample uniformly in three dimensions: rotation angle, inverse radial distance, and vertical elevation. The use of multiperspective panoramas eliminates the limited overlap present in the original input images and, thus, problems as in conventional multibaseline stereo can be avoided. Our approach differs from stereo matching of single-perspective panoramic images taken from different locations, where the epipolar constraints are sine curves. For our multiperspective panoramas, the epipolar geometry, to the first order approximation, consists of horizontal lines. Therefore, any traditional stereo algorithm can be applied to multiperspective panoramas with little modification. In this paper, we describe two reconstruction algorithms. The first is a cylinder sweep algorithm that uses a small number of resampled multiperspective panoramas to obtain dense 3D reconstruction. The second algorithm, in contrast, uses a large number of multiperspective panoramas and takes advantage of the approximate horizontal epipolar geometry inherent in multiperspective panoramas. It comprises a novel and efficient 1D multibaseline matching technique, followed by tensor voting to extract the depth surface. Experiments show that our algorithms are capable of producing comparable high quality depth maps which can be used for applications such as view interpolation.     

由抖动图象序列建立自然环境三维真实感全景模型

由真实环境中的现场图象进行三维环境建模是目前国际上研究的热点问题。本文依据合理的运动模型，提出和实现了由包含抖动的摄像机运动下的图象序列建立３Ｄ环境全景模型的两步法。首先通过运动滤波和运动分解获得运动稳定的图象序列，然后采用无特征提取的时空纹理方向精确估计、深度边界确定和遮挡恢复算法，建立全局自然景物的真实感三维环境模型。提出了２种三维全景图象的表示方法，即非阵列方式深度分层区域表示和阵列方式的深度分层布景表示，可用于机器人全局定位的自然路标提取和真实环境虚拟再现的图象合成。该研究推广和结合了外极面图象的方法和全景图象的方法，放宽了对运动的要求，从而可使该种方法适用于室外颠簸的道路环境。和现有运动分层方法相比，避免了该类方法迭代过程中的局部最小化问题，并具有计算和存储效率高，适应性强，算法鲁棒性好的优点。     

Large-scale 3D Semantic Mapping Using Stereo Vision

In recent years, there have been a lot of interests in incorporating semantics into simultaneous localization and mapping (SLAM) systems. This paper presents an approach to generate an outdoor large-scale 3D dense semantic map based on binocular stereo vision. The inputs to system are stereo color images from a moving vehicle. First, dense 3D space around the vehicle is constructed, and the motion of camera is estimated by visual odometry. Meanwhile, semantic segmentation is performed through the deep learning technology online, and the semantic labels are also used to verify the feature matching in visual odometry. These three processes calculate the motion, depth and semantic label of every pixel in the input views. Then, a voxel conditional random field (CRF) inference is introduced to fuse semantic labels to voxel. After that, we present a method to remove the moving objects by incorporating the semantic labels, which improves the motion segmentation accuracy. The last is to generate the dense 3D semantic map of an urban environment from arbitrary long image sequence. We evaluate our approach on KITTI vision benchmark, and the results show that the proposed method is effective.     

丝路文化虚拟体验中的多视角立体重建技术研究

丝路文化是联系一带一路战略的重要纽带,其传承意义重大,但是由于历史地理原因,丝路文化中代表性的历史遗产分散或损坏,难以有效地呈现,因此,本文面向丝路文化的虚拟展示与数字化,提出并实现了基于虚拟现实技术的丝路文化传承平台,通过历史遗迹复原以及基于图像的三维重建,还原了丝路文化中重要节点宁夏固原有关的历史遗迹、文物和事件....     

Attention-Aware Disparity Control in interactive environments

Our paper introduces a novel approach for controlling stereo camera parameters in interactive 3D environments in a way that specifically addresses the interplay of binocular depth perception and saliency of scene contents. Our proposed Dynamic Attention-Aware Disparity Control (DADC) method produces depth-rich stereo rendering that improves viewer comfort through joint optimization of stereo parameters. While constructing the optimization model, we consider the importance of scene elements, as well as their distance to the camera and the locus of attention on the display. Our method also optimizes the depth effect of a given scene by considering the individual user’s stereoscopic disparity range and comfortable viewing experience by controlling accommodation/convergence conflict. We validate our method in a formal user study that also reveals the advantages, such as superior quality and practical relevance, of considering our method.     

3D Scene Reconstruction from Multiple Spherical Stereo Pairs

We propose a 3D environment modelling method using multiple pairs of high-resolution spherical images. Spherical images of a scene are captured using a rotating line scan camera. Reconstruction is based on stereo image pairs with a vertical displacement between camera views. A 3D mesh model for each pair of spherical images is reconstructed by stereo matching. For accurate surface reconstruction, we propose a PDE-based disparity estimation method which produces continuous depth fields with sharp depth discontinuities even in occluded and highly textured regions. A full environment model is constructed by fusion of partial reconstruction from spherical stereo pairs at multiple widely spaced locations. To avoid camera calibration steps for all camera locations, we calculate 3D rigid transforms between capture points using feature matching and register all meshes into a unified coordinate system. Finally a complete 3D model of the environment is generated by selecting the most reliable observations among overlapped surface measurements considering surface visibility, orientation and distance from the camera. We analyse the characteristics and behaviour of errors for spherical stereo imaging. Performance of the proposed algorithm is evaluated against ground-truth from the Middlebury stereo test bed and LIDAR scans. Results are also compared with conventional structure-from-motion algorithms. The final composite model is rendered from a wide range of viewpoints with high quality textures.     

A 3D Imaging Framework Based on High-Resolution Photometric-Stereo and Low-Resolution Depth

This paper introduces a 3D imaging framework that combines high-resolution photometric stereo and low-resolution depth. Our approach targets imaging scenarios based on either macro-lens photography combined with focal stacking or a large-format camera that are able to image objects with more than 600 samples per mm $^2$ . These imaging techniques allow photometric stereo algorithms to obtain surface normals at resolutions that far surpass corresponding depth values obtained with traditional approaches such as structured-light, passive stereo, or depth-from-focus. Our work offers two contributions for 3D imaging based on these scenarios. The first is a multi-resolution, patched-based surface reconstruction scheme that can robustly handle the significant resolution difference between our surface normals and depth samples. The second is a method to improve the initial normal estimation by using all the available focal information for images obtained using a focal stacking technique.     

Panoramic Depth Imaging: Single Standard Camera Approach

In this paper we present a panoramic depth imaging system. The system is mosaic-based which means that we use a single rotating camera and assemble the captured images in a mosaic. Due to a setoff of the camera's optical center from the rotational center of the system we are able to capture the motion parallax effect which enables stereo reconstruction. The camera is rotating on a circular path with a step defined by the angle, equivalent to one pixel column of the captured image. The equation for depth estimation can be easily extracted from the system geometry. To find the corresponding points on a stereo pair of panoramic images the epipolar geometry needs to be determined. It can be shown that the epipolar geometry is very simple if we are doing the reconstruction based on a symmetric pair of stereo panoramic images. We get a symmetric pair of stereo panoramic images when we take symmetric pixel columns on the left and on the right side from the captured image center column. Epipolar lines of the symmetrical pair of panoramic images are image rows. The search space on the epipolar line can be additionaly constrained. The focus of the paper is mainly on the system analysis. Results of the stereo reconstruction procedure and quality evaluation of generated depth images are quite promissing. The system performs well for reconstruction of small indoor spaces. Our finall goal is to develop a system for automatic navigation of a mobile robot in a room.     

任意视角的多视图立体匹配系统

为了得到高精度深度图,通过特征点提取与匹配、计算基础矩阵F、引导互匹配、捆集调整等一系列技术,求出每幅视图在同一空间坐标系下的高精度摄像机矩阵P.为任意角度的多个视图建立可扩展的主框架,不需要图像矫正.基于图像坐标和自动计算出的视差范围建立三维网络,为每条边分配适当的权值,把多视图立体匹配问题转化为网络最大流问题.算法保证了高准确率和平滑性,实验结果表明此方法适用于三维模型重建.     

An automatic 2D to 3D video conversion approach based on RGB-D images

3D movies/videos have become increasingly popular in the market; however, they are usually produced by professionals. This paper presents a new technique for the automatic conversion of 2D to 3D video based on RGB-D sensors, which can be easily conducted by ordinary users. To generate a 3D image, one approach is to combine the original 2D color image and its corresponding depth map together to perform depth image-based rendering (DIBR). An RGB-D sensor is one of the inexpensive ways to capture an image and its corresponding depth map. The quality of the depth map and the DIBR algorithm are crucial to this process. Our approach is twofold. First, the depth maps captured directly by RGB-D sensors are generally of poor quality because there are many regions missing depth information, especially near the edges of objects. This paper proposes a new RGB-D sensor based depth map inpainting method that divides the regions with missing depths into interior holes and border holes. Different schemes are used to inpaint the different types of holes. Second, an improved hole filling approach for DIBR is proposed to synthesize the 3D images by using the corresponding color images and the inpainted depth maps. Extensive experiments were conducted on different evaluation datasets. The results show the effectiveness of our method.

     

Automatic 3D object segmentation in multiple views using volumetric graph-cuts

We propose an algorithm for automatically obtaining a segmentation of a rigid object in a sequence of images that are calibrated for camera pose and intrinsic parameters. Until recently, the best segmentation results have been obtained by interactive methods that require manual labelling of image regions. Our method requires no user input but instead relies on the camera fixating on the object of interest during the sequence. We begin by learning a model of the object’s colour, from the image pixels around the fixation points. We then extract image edges and combine these with the object colour information in a volumetric binary MRF model. The globally optimal segmentation of 3D space is obtained by a graph-cut optimisation. From this segmentation an improved colour model is extracted and the whole process is iterated until convergence.     

Vision-based global localization for mobile robots with hybrid maps of objects and spatial layouts

This paper presents a novel vision-based global localization that uses hybrid maps of objects and spatial layouts. We model indoor environments with a stereo camera using the following visual cues: local invariant features for object recognition and their 3D positions for object pose estimation. We also use the depth information at the horizontal centerline of image where the optical axis passes through, which is similar to the data from a 2D laser range finder. This allows us to build our topological node that is composed of a horizontal depth map and an object location map. The horizontal depth map describes the explicit spatial layout of each local space and provides metric information to compute the spatial relationships between adjacent spaces, while the object location map contains the pose information of objects found in each local space and the visual features for object recognition. Based on this map representation, we suggest a coarse-to-fine strategy for global localization. The coarse pose is estimated by means of object recognition and SVD-based point cloud fitting, and then is refined by stochastic scan matching. Experimental results show that our approaches can be used for an effective vision-based map representation as well as for global localization methods.     

Inertial Sensed Ego‐motion for 3D Vision

Inertial sensors attached to a camera can provide valuable data about camera pose and movement. In biological vision systems, inertial cues provided by the vestibular system are fused with vision at an early processing stage. In this article we set a framework for the combination of these two sensing modalities. Cameras can be seen as ray direction measuring devices, and in the case of stereo vision, depth along the ray can also be computed. The ego‐motion can be sensed by the inertial sensors, but there are limitations determined by the sensor noise level. Keeping track of the vertical direction is required, so that gravity acceleration can be compensated for, and provides a valuable spatial reference. Results are shown of stereo depth map alignment using the vertical reference. The depth map points are mapped to a vertically aligned world frame of reference. In order to detect the ground plane, a histogram is performed for the different heights. Taking the ground plane as a reference plane for the acquired maps, the fusion of multiple maps reduces to a 2D translation and rotation problem. The dynamic inertial cues can be used as a first approximation for this transformation, allowing a fast depth map registration method. They also provide an image independent location of the image focus of expansion and center of rotation useful during visual based navigation tasks. © 2004 Wiley Periodicals, Inc.     

深度学习单目深度估计研究进展

单目深度估计是从单幅图像中获取场景深度信息的重要技术,在智能汽车和机器人定位等领域应用广泛,具有重要的研究价值。随着深度学习技术的发展,涌现出许多基于深度学习的单目深度估计研究,单目深度估计性能也取得了很大进展。本文按照单目深度估计模型采用的训练数据的类型,从3个方面综述了近年来基于深度学习的单目深度估计方法:基于单图像训练的模型、基于多图像训练的模型和基于辅助信息优化训练的单目深度估计模型。同时,本文在综述了单目深度估计研究常用数据集和性能指标基础上,对经典的单目深度估计模型进行了性能比较分析。以单幅图像作为训练数据的模型具有网络结构简单的特点,但泛化性能较差。采用多图像训练的深度估计网络有更强的泛化性,但网络的参数量大、网络收敛速度慢、训练耗时长。引入辅助信息的深度估计网络的深度估计精度得到了进一步提升,但辅助信息的引入会造成网络结构复杂、收敛速度慢等问题。单目深度估计研究还存在许多的难题和挑战。利用多图像输入中包含的潜在信息和特定领域的约束信息,来提高单目深度估计的性能,逐渐成为了单目深度估计研究的趋势。     

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。     

A multicamera setup for generating stereo panoramic video

Traditional visual communication systems convey only two-dimensional (2-D) fixed field-of-view (FOV) video information. The viewer is presented with a series of flat, nonstereoscopic images, which fail to provide a realistic sense of depth. Furthermore, traditional video is restricted to only a small part of the scene, based on the director's discretion and the user is not allowed to "look around" in an environment. The objective of this work is to address both of these issues and develop new techniques for creating stereo panoramic video sequences. A stereo panoramic video sequence should be able to provide the viewer with stereo vision at any direction (complete 360-degree FOV) at video rates. In this paper, we propose a new technique for creating stereo panoramic video using a multicamera approach, thus creating a high-resolution output. We present a setup that is an extension of a previously known approach, developed for the generation of still stereo panoramas, and demonstrate that it is capable of creating high-resolution stereo panoramic video sequences. We further explore the limitations involved in a practical implementation of the setup, namely the limited number of cameras and the nonzero physical size of real cameras. The relevant tradeoffs are identified and studied.     

A 1D approach to correlation-based stereo matching

In stereovision, indices allowing pixels of the left and right images to be matched are basically one-dimensional features of the epipolar lines. In some situations, these features are not significant or cannot be extracted from the single epipolar line. Therefore, many techniques use 2D neighbourhoods to increase the available information. In this paper, we discuss the systematic use of 2D neighbourhoods for stereo matching. We propose an alternative approach to stereo matching using multiple 1D correlation windows, which yields a semi-dense disparity map and an associated confidence map. A particular technique derived from this approach — using fuzzy filtering and a basic decision rule — is compared to about 80 other methods on the Middlebury image datasets [1]. Results are first presented in the framework of the Middlebury website, then on the Receiver Operating Characteristics (ROC) evaluation [2] and, finally, on stereo image pairs of slanted surfaces. We show that a 1D correlation window is sufficient to provide correct matchings in most cases.     

基于视差平面分割的移动机器人障碍物地图构建方法

作为自主移动机器人地表障碍物探测（GPOD）技术的一部分,提出了一种利用双目摄像机的视差图像 获取信息来构建机器人前方障碍物栅格地图的方法. 该方法融合了3 维立体视觉技术以及2 维图像处理技术,前者 依据视差图的直方图信息对视差图像进行自适应平面分割,把每个平面看作是3 维场景中的实物切片进而提取障碍 物3 维信息,后者通过计算各平面上的障碍物信息曲线来提取障碍物信息,把立体视觉数据从视差图像空间变换到 2 维的障碍物地图空间. 给出了该方法构建障碍物地图的整体过程,试验结果证明了该算法的有效性和精确性.     

Video 2‐D—to—3‐D conversion based on hybrid depth cueing

Abstract— With the maturation of three‐dimensional (3‐D) technologies, display systems can provide higher visual quality to enrich the viewer experience. However, the depth information required for 3‐D displays is not available in conventional 2‐D recorded contents. Therefore, the conversion of existing 2‐D video to 3‐D video becomes an important issue for emerging 3‐D applications. This paper presents a system which automatically converts 2‐D videos to 3‐D format. The proposed system combines three major depth cues: the depth from motion, the scene depth from geometrical perspective, and the fine‐granularity depth from the relative position. The proposed system uses a block‐based method incorporating a joint bilateral filter to efficiently generate visually comfortable depth maps and to diminish the blocky artifacts. By means of the generated depth map, 2‐D videos can be readily converted into 3‐D format. Moreover, for conventional 2‐D displays, a 2‐D image/video depth perception enhancement application is also presented. With the depth‐aware adjustment of color saturation, contrast, and edge, the stereo effect of the 2‐D content can be enhanced. A user study on subjective quality shows that the proposed method has promising results on depth quality and visual comfort.     

一种将羽毛球比赛的2D视频转换到3D视频的算法

文中提出一种羽毛球比赛的2D视频转换到3D视频的算法。在这类视频中,前景是最受关注的部分,准确地从背景中提取出前景对象是获取深度图的关键。文中采用一种改进的图割算法来获取前景,并根据场景结构构建背景深度模型,获取背景深度图;在背景深度图的基础上,根据前景与镜头之间的距离关系为前景对象进行深度赋值,从而得到前景深度图。然后,融合背景深度图和前景深度图,得到完整的深度图。最后,通过基于深度图像的虚拟视点绘制技术DIBR来获取用于3D显示的立体图像对。实验结果表明,最终生成的立体图像对具有较好的3D效果。