新的单目立体视觉的视差图的获得方法

在立体视觉中,视差间接反映物体的深度信息,视差计算是深度计算的基础。常见的视差计算方法研究都是面向双目立体视觉,而双焦单目立体视觉的视差分布不同于双目视差,具有沿极线辐射的特点。针对双焦单目立体视觉的特点,提出了一种单目立体视差的计算方法。对于计算到的初步视差图,把视差点分类为匹配计算点和误匹配点。通过均值偏移向量（Mean Shift）算法,实现了对误匹配点依赖于匹配点和图像分割的视差估计,最终得到致密准确的视差图。实验证明,这种方法可以通过双焦立体图像对高效地获得场景的视差图。     

人眼立体视觉空间与虚拟三维空间的几何映射关系理论分析

虚拟三维空间是现实世界的数字化三维空间,而人眼立体视觉空间则是人眼视觉系统对于现实世界或虚拟世界所形成的三维立体构象。传统上人眼直接观察现实世界,确立了人眼立体视觉空间与现实世界之间的几何对应关系。而人眼立体视觉空间与虚拟三维空间是否也存在对应的几何关系？以人眼立体视觉和虚拟三维场景为研究对象,根据双目视差原理,论述了人眼立体视觉的几何模型及视觉三维模型的表示形式。通过分析虚拟空间三维点元、屏幕视差及网膜视差等三者之间的内在几何关系,利用矩阵代数建立了虚拟空间与视觉空间之间的几何映射关系。这一映射关系表明视觉三维模型与虚拟三维模型之间存在一一对应关系,也反映了人眼视觉系统在虚拟空间中的可测量性质。本文创新性之处在于得到了视觉三维模型的完整表示,突破了传统上立体显示的定性感知,提供了定量分析的基础。这一工作对于在虚拟空间中的立体体验、虚拟交互以及立体测量等实践活动具有一定理论参考价值。     

结合立体视觉舒适度的立体图像显著性检测

针对先前的立体图像显著性检测模型未充分考虑立体视觉舒适度和视差图分布特征对显著区域检测的影响,提出了一种结合立体视觉舒适度因子的显著性计算模型.该模型在彩色图像显著性提取中,首先利用SLIC算法对输入图像进行超像素分割,随后进行颜色相似区域合并后再进行二维图像显著性计算;在深度显著性计算中,首先对视差图进行预处理;然后基于区域对比度进行显著性计算;最后,结合立体视觉舒适度因子对二维显著图和深度显著图进行融合,得到立体图像显著图.在不同类型立体图像上的实验结果表明,该模型获得了85%的准确率和78%的召回率,优于现有常用的显著性检测模型,并与人眼立体视觉注意力机制保持良好的一致性.     

基于人眼视觉原理的虚拟现实显示模型

针对虚拟现实领域现有的显示模块无法提供满足人眼视觉规律的立体视觉效果这一问题,提出一种基于斜交视锥体立体摄像机模型的虚拟现实(VR)立体视觉解决方案。首先,通过研究人眼视域模型和人眼双目提取深度信息原理,建立双目视差数学模型;其次,使用工业引擎3DVIA Studio作为实验平台,依靠VSL编程语言进行画面分屏处理,通过设立父子关系,设置视觉交互模块,搭建VR交互立体摄像机结构;接着,通过建立点云模型,量化物体立体感观,探讨多种VR立体摄像机模型,从深度信息显示效果和畸变特性出发分析各个模型优缺点,逐步进行优化。分别建立了中心轴平行的正交视锥体摄像机模型、中心轴于观影距离处相交的正交视锥体模型,通过优化摄像机视锥体结构,形成斜交视锥体摄像机模型。最后使用3DVIA Studio进行实验,代入具体数据进行投影变换,结果显示斜交视锥体摄像机模型在保证深度信息显示效果的前提下很好地消除了畸变,能够提供优秀的视觉体验效果。     

实时立体视觉系统中的深度映射

目的 为减少立体图像中由于水平视差过大引起的视觉疲劳。针对实时渲染的立体视觉系统,给出了一种非均匀深度压缩方法。方法 该方法在单一相机空间内,通过不同的投影变换矩阵生成双眼图像,水平视差由投影变换来控制。为减少深度压缩造成的模型变形而带来的瑕疵,将不同深度区域内物体施以不同的压缩比例;将相机轴距表示为深度的连续函数,通过相机轴距推导出在单一相机空间内获取双眼图像的坐标变换,将深度压缩转换为模型的坐标变换,从而保证压缩比例的连续变化。结果 实验结果表明,该方法能有效提高立体图像的质量。结论 该方法简单、高效,可应用于游戏、虚拟现实等实时立体视觉系统。     

多目立体视觉在工业测量中的应用研究

以双目立体视觉系统为基础，提出了一种多目立体视觉测量的方法。该方法利用摄像机旋转来获取多幅图像，利用这些立体图像对中对应特征点的视差得到物体的深度信息，达到多目测量的目的，通过实验并与双目立体视觉相比较，验证了其在摄像机标定和提高测量结果精度方面的优越性。     

一种测量高光物体的双目Helmholtz立体视觉方法

提出了一种基于迭代动态规划的双目Helmholtz立体视觉算法,并将其应用于高光物体的测量。算法首先对获取的Helmholtz图像对作图像校正,然后确定扫描线的端点,最后设计了一种迭代动态规划方法建立匹配获取视差图,从而恢复出高光物体的表面深度。实验采用光线跟踪方法获取带有高光的双目Helmholtz图像,视差图结果表明该方法能够有效地恢复出高光物体的深度信息。     

基于单目和双目视觉信息的全参考立体图像质量评价模型

在立体图像质量评价领域,有效地模拟人类视觉系统对图像质量进行评价具有重要意义,考虑到人眼的视觉感知特性,基于单目和双目视觉信息构建一种立体图像质量评价模型MB-FR-SIQA。采用基于结构相似性的立体视差算法得到参考和失真立体图像的视差矩阵,结合Gabor能量响应图、显著性图和视差矩阵生成中间视图,并优化左右眼加权系数计算方法,以提高生成中间视图的准确性。分别利用单目图像和中间视图提取单目和双目视觉信息,计算单目质量分数和双目质量分数,并融合得到立体图像的质量分数,达到评价立体图像质量的目的。实验结果表明,MB-FR-SIQA模型在LIVE-I数据库上具有较高的预测精度,其斯皮尔曼等级相关系数、皮尔森线性相关系数、均方根误差分别为0.945、0.951、5.318,且预测的质量分数符合人类主观评估。     

面向双目立体视觉的迭代式局部颜色校正

在双目立体视觉中,由于相机参数设置差异、环境光照变化、拍摄物体表面非理想漫反射等因素,拍摄获得的立体图像对可能存在颜色差异,进而降低视差计算的准确性.针对这一问题,提出一种面向双目立体视觉的迭代式局部颜色校正方法.首先使用Meanshift算法以不同粒度对2幅图像进行分割,并基于SIFT特征匹配、区域分布及颜色差异初步建立2幅图像中物体间的对应关系;然后使用加权局部颜色校正方法,逐区域进行颜色校正,由于双目图像中物体遮挡范围不同,初步的区域对应存在误差,因此利用校正的双目图像计算视差图,基于视差图像对之间稠密的像素对应,优化立体图像对的区域分割,建立更准确的对应关系,并再次进行颜色校正;迭代地进行立体匹配、优化图像区域对应和颜色校正,直至获得最佳的立体匹配结果.与已有颜色校正方法对基准测试图像集的处理结果表明,文中方法可以有效地提升立体图像对的颜色相似度,提高立体匹配视差结果的准确性.     

基于视差重映射的立体图像视觉舒适度提升

目的 针对人眼观看立体图像内容可能存在的视觉不舒适性,基于视差对立体图像视觉舒适度的影响,提出了一种结合全局线性和局部非线性视差重映射的立体图像视觉舒适度提升方法。方法 首先,考虑双目融合限制和视觉注意机制,分别结合空间频率和立体显著性因素提取立体图像的全局和局部视差统计特征,并利用支持向量回归构建客观的视觉舒适度预测模型作为控制视差重映射程度的约束;然后,通过构建的预测模型对输入的立体图像的视觉舒适性进行分析,就欠舒适的立体图像设计了一个两阶段的视差重映射策略,分别是视差范围的全局线性重映射和针对提取的潜在欠舒适区域内视差的局部非线性重映射;最后,根据重映射后的视差图绘制得到舒适度提升后的立体图像。结果 在IVY Lab立体图像舒适度测试库上的实验结果表明,相较于相关有代表性的视觉舒适度提升方法对于欠舒适立体图像的处理结果,所提出方法在保持整体场景立体感的同时,能更有效地提升立体图像的视觉舒适度。结论 所提出方法能够根据由不同的立体图像特征构建的视觉舒适度预测模型来自动实施全局线性和局部非线性视差重映射过程,达到既改善立体图像视觉舒适度、又尽量减少视差改变所导致的立体感削弱的目的,从而提升立体图像的整体3维体验。     

Interactive stereoscopic displays

Interactive stereo displays allow for the existence of a natural interaction between the user and the stereo images depicted on the display. In the type of display discussed here, this interaction takes the form of tracking the user's head and hand/arm position. Sensing the user's head position allows for the creation of motion parallax information, an immersive depth cue that can be added to the binocular parallax already present in the display. Sensing the user's hand or arm position allows the user to manipulate the spatial attributes of virtual objects and scenes presented on the display, which can enhance spatial reasoning. Moreover, allowing the user to manipulate virtual objects may permit the creation of a sense of spatial relations among elements in the display via proprioception, which may augment the two parallax cues. The congruence among binocular parallax, motion parallax, and proprioception should increase the sense of depth in the display and increase viewing comfort, as well as enhance the ability of our intuitive reasoning system to make reasoned sense out of the perceptual information. These advantages should make interactive stereo displays, which may be classified as a form of cognitive enhancement display, the display of choice in the future.     

Impact of parallax and interpupillary distance on size judgment performances of virtual objects in stereoscopic displays

Effective interactions in both real and stereoscopic environments require accurate perceptions of size and position. This study investigated the effects of parallax and interpupillary distance (IPD) on size perception of virtual objects in widescreen stereoscopic environments. Twelve participants viewed virtual spherical targets displayed at seven different depth positions, based on seven parallax levels. A perceptual matching task using five circular plates of different sizes was used to report the size judgment. The results indicated that the virtual objects were perceived as larger and smaller than the corresponding theoretical sizes, respectively, in negative and positive parallaxes. Similarly, the estimates from participants with small IPDs were greater than the predicted estimates. The findings of this study are used to explain human factor issues such as the phenomenon of inaccurate depth judgments in virtual environments, where compression is widely reported, especially at farther egocentric distances. Furthermore, a multiple regression model was developed to describe how the size was affected by parallax and IPD.

Practitioner Summary: The study investigates the effects of parallax and interpupillary distance on size perception of virtual targets in a stereoscopic environment. Virtual objects were perceived as larger in negative and smaller in positive parallax. Also, size estimates were greater than the theoretical sizes for participants with smaller IPD. A multiple-regression model explains the impact of parallax and measured IPD. Abbreviations IPD interpupillary distance

VR virtual eality

HMD head mounted-displays

2AFC two-alternative forced choice

IOD interocular distance

PD pupillary distance

ANOVA analysis of variance

     

虚拟环境中的视觉感知

通过对现实世界,平面图像的视觉成象方式的比较,指出虚拟环境成象方式是二者相结合的产物。在论述虚拟环境所提供的视差对三维深度感知重要性的同时,结合视觉生理特点,分析总结了虚拟环境所具有的缺陷,即违背了视标所发出的入瞳光线的散开度与双眼所处的状态应该相辅相成这一规律,提出了虚拟环境给队造成头晕、眼睛疲劳等不良反应的根源是虚拟对象容易使双眼的辐辏和调节功能发生冲突。最后列举并分析了刷新延迟、分辩率、瞳距等各种可能给观察者带来不良反应的诸多因素。     

Virtual toed‐in camera method to eliminate parallax distortions of stereoscopic images for stereoscopic displays

Abstract— The parallax images captured by a toed‐in camera have positive and negative horizontal parallax for stereoscopic displays, but they also have parallax distortions including horizontal and vertical distortions. The virtual toed‐in camera method can eliminate the horizontal distortion and reduce the vertical distortion. The parallax formulae for the corrected‐parallax images were deduced. The experiments were carried out and the results proved that the method was accurate. Good stereoscopic images without horizontal distortion and with little vertical distortion were presented on an autostereoscopic display.     

Human fusion range for polarized 3D display

If discrepancy between accommodation and convergence caused by a stereoscopic display exceeds fusion range of human eyes, viewers will see ghosting image, which leads to the loss of correct depth information and even causes severe visual fatigue. In this paper, an experiment aiming to investigate the binocular fusion range is conducted for a polarized 3D display. Two experimental trials are arranged to examine two aspects of fusion range including outward depth and inward depth. 3D modeling software is used to generate the test stereoscopic image pairs, which vary in depth by adjusting the separation between the virtual cameras. Angular parallax corresponding to the limit of fusion range is obtained by determining critical point of ghosting images. The experimental results show deviation between theoretical fusion range calculated by formula and experimental one.?0.223° to 0.275° represent critical fusion range for the polarized 3D display to avoid ghosting images.     

3D Time-Lapse Reconstruction from Internet Photos

Given an Internet photo collection of a landmark, we compute a 3D time-lapse video sequence where a virtual camera moves continuously in time and space. While previous work assumed a static camera, the addition of camera motion during the time-lapse creates a very compelling impression of parallax. Achieving this goal, however, requires addressing multiple technical challenges, including solving for time-varying depth maps, regularizing 3D point color profiles over time, and reconstructing high quality, hole-free images at every frame from the projected profiles. Our results show photorealistic time-lapses of skylines and natural scenes over many years, with dramatic parallax effects.     

Homography-based depth recovery with descent images

In planetary landing exploration task, the images captured by the landing camera are nearly along optical axis which results in multi-resolution images of same terrain surface. Recovering the surface shape of landing terrain from descent imagery is of great value for lander to choose safe landing area. In this paper, a homography-based depth recovery method with descent images is addressed. At first, the parallax and scale change in descent images are analyzed. Second, the camera motion is optimized with SIFT features correspondence constraints. For dense depth recovery, a set of virtual parallel planes is assumed to slice the terrain and each plane induces a homography to warp back the second image to first image plane. Zero-normalized cross-correlation score is chosen to compute the correlation score and the correlation curve is smoothed by two Gaussian filters. The depth for each pixel is determined by the plane which has highest correlation value. At the end, some experiments are conducted, including different correlation computation, depth recovery with different terrain, and the error tests. The results show that the discussed method is feasible to recover the depth information overall.     

Can movement parallax compensate lacking stereopsis in spatial explorative search tasks?

In an explorative task subjects were asked to track a line presented in combination with other confusing lines on a 3D display. 24 subjects performed the explorative task under three depth cue conditions which were: stereopsis, movement parallax or a combination of stereopsis and movement parallax (stereo and parallax).

Task performance was assessed by recording completion time and correctness of answer.

The combination of stereopsis and movement parallax gave rise to highest percentage of correct answers (84%), followed by the condition where movement parallax was the only depth cue (80%). Percentage of correct answers in both conditions was found to be significantly higher than it was in the condition with stereopsis (60%).

Mean completion time was about the same in the stereo and parallax condition and the stereopsis condition. Compared to the movement parallax condition both means were significantly decreased.

Accuracy in completing the task was found to be similar in 3D displays enabling movement parallax and in 3D displays presenting stereoscopic images.     

Dual layered display that presents auto‐stereoscopic 3D images to multiple viewers in arbitrary positions

Dual layered display or also called tensor display that consists of two panels in a stack can present full‐parallax 3D images with high resolution and continuous motion parallax by reconstructing corresponding light ray field within a viewing angle. The depth range where the 3D images can be displayed with reasonable resolution, however, is limited around the panel stack. In this paper, we propose a dual layered display that can present stereoscopic images to multiple viewers located at arbitrary positions in observer space with high resolution and large depth range. Combined with the viewer tracking system, the proposed method provides a practical way to realize high‐resolution large‐depth auto‐stereoscopic 3D display for multiple observers without restriction on the observer position and the head orientation.