2022-11-26 19:39:55

Advances in Image and Video Technology

Abstract.With the advance of 3D digitization and rendering technologies, interactive virtual exhibition can now be realized for applications such as virtual museum, virtual showcase, and virtual mall. There are two major approaches to implementing a 3D interactive virtual exhibition application. One approach is the geometry-based approach, which reconstructs geometric models for 3D objects by using laser scanners or other 3D digitization equipments. Another approach is the imagebased approach, which renders the 3D object directly using a large set of pre-acquired images without reconstructing geometric models. While the geometry-based approach provides better interaction and smaller data size, its cost-effectiveness is not as good as the image-based approach. We have developed a new image-based approach to 3D interactive virtual exhibition based on a technique named augmented panoramas. With augmented panoramas, the 3D exhibition space is represented by panoramas and can be augmented by the 3D objects to be exhibited, which can be represented either by geometric models or by object movies. Compared with other image-based techniques, such as light field rendering, object movies have the advantage of being easier in image acquisition and rendering. Here, a major challenge for augmented panoramas is how to integrate two sources of 2D images, the panoramas and the object movies, in a 3D-consistent way. In Taiwan, with the support of National Digital Archives Program, we have designed and implemented a 3D stereoscopic kiosk system for virtually exhibiting 3D artifacts in the National Palace Museum, the National Historical Museum, and the Museum of the Institute of History and Philology, Academia Sinica. Also, we have built a few other interactive display systems, which will be shown in this presentation. For example, we shall show the Magic Crystal Ball, which allows the user to see a virtual object appearing inside a transparent ball and to rotate the virtual object by barehanded interaction. Our goal is to transform different concepts from movies and fiction into the development of a new medium for the users to access multimedia in an intuitive, imaginative and playful manner.

Out-of-Order Execution for Avoiding Head-of-Line Blocking in Remote 3D Graphics John Stavrakakis and Masahiro Takastuka

Abstract. Remote 3D graphics can become both process and network intensive. The Head-of-Line Blocking(HOLB) problem exists for an ordered stream protocol such as TCP. It withholds any available data from the application until the proper ordered segment arrives. The HOLB problem will cause the processor to have unnecessary idle time and non-uniform load patterns. In this paper we evaluate how the performance of an immediate mode remote 3D graphics system is affected by the HOLB and how the out-of-order execution can improve the performance. Keyword: Distributed rendering, Network graphics, Load balancing.

1 Introduction Interactive multimedia applications over a network demand both realtime delivery and excellent response time for a high quality end user experience. Networked applications using 3D graphics have difficulties ensuring this level of quality is maintained. The biggest problem is the sheer volume of network traffic they would generate. For audio and video multimedia, this is greatly reduced through appropriate down-sampling and discardment of perceptually unimportant data. In addition, they are able to exploit lossy network protocols that can continue without all the data[1]. This scenario is contrast to that of 3D graphics, as the the majority of data be preserved correctly which would otherwise cause severe artifacts to appear in the rendered images. Specifically, the complexity of graphics data creates greater network traffic and thus making it diffi- cult to maintain desirable rendering frame rates. Networked 3D graphics has been well researched for high network utilisation, compression [2][3] and several techniques in reducing computational load [4]. Despite the attention, another bottleneck existing in the network graphics systems occurs between the passing of data from the network layer to the rendering engine. As the data is received from the network in the form of segments (a one-to-one correspondence to a packet sent/delivered), the segments are held in buffers until reassembly of the application data can take place. Following the reassembly of fragments to a segment, the segment will need to meet an ordering requirement prior to being passed on for rendering. Such a process is typically handled between decoupled networking and rendering modules within the system. This problem is also known as Head-of-Line Blocking, and exists in the Transmission Control Protocol[5] (TCP) protocol. It occurs when a TCP segment is lost and asubsequent TCP segment arrives out of order. The subsequent segment is held until the first TCP segment is retransmitted and arrives at the receiver[6]. This is not a problem for most software using TCP, as the requirement is not realtime and operation order is essential. Immediate mode graphics works differently; each command is simply defining the rendered scene, this requires no order until the drawing takes place. As such, 3D graphics is able to avoid the HOLB problem as its operations need not execute in order. The advantage of doing so allows the the graphics processor to avoid idling when data is available to execute. This paper investigates the avoidance of the HOLB for a subset of applications using immediate mode 3D graphics. The following section will briefly introduce motivation for remote 3D graphics and currently available systems. Section 3 will detail the out of order nature in 3D graphics and its theoretical impact. We follow with experimental methods and results to address HOLB, and finally conclude with a brief summary of our findings and future work.

2 Remote 3D Visuali



主题演讲基于图像的方法交互式3 d虚拟展览。张揖平挂图像和视觉实验室研究所网络和多媒体计算机科学与信息工程系国立台湾大学联合研究。与3 d数字化的推进和渲染技术,交互式虚拟展览等应用程序现在可以实现虚拟博物馆,虚拟展示和虚拟商场。

主要有两种方法实现三维交互式虚拟展示的应用程序。一种方法是几何投影的方法,重新构建几何模型3 d对象通过使用激光扫描仪或其他3 d数字化设备。除此之外的另一种方法是,直接呈现3 d对象使用一套大型的pre-acquired图像不需要重建几何模型。而几何投影方法提供了更好的互动和较小的数据大小,其成本效益不如基于图像的方法。

我们已经开发出一种新的基于图像的三维交互式虚拟展览的方法,基于全景图技术命名的增强。与3 d增强全景照片展览空间是由全景照片,可以表现出增强的3 d对象,可以通过几何模型或对象表示出电影中的一些部分。与其它基于映像的技术相比,光场渲染、对象等电影的优点是容易在图像采集和渲染。全景图,增强的一个主要的挑战是如何将两个来源的2 d图像对象的全貌和电影,3 d-consistent。在台湾,国家数字档案馆项目的支持下,我们设计并实现了一个3 d立体成像系统几乎表现出三维工件在国立故宫博物院,国家历史博物馆,博物馆的历史和语言学研究所等中央研究院。同时,我们建立了一些其他互动显示系统,将本课程所示。例如,我们将显示魔法水晶球,它允许用户查看虚拟对象出现在一个透明的球和旋转赤手空拳的交互的虚拟对象。我们的目标是把不同的概念从电影和小说的发展新媒体的用户访问多媒体以直观的、富有想象力和有趣的方式。

无序执行避免连接阻塞3 d图形转换和显示

抽象的远程3 d图形可以成为过程和网络带宽。Head-of-Line阻塞(HOLB)问题的存在等有序流协议TCP。它保留任何可用的数据从应用程序到正确的命令部分到达。HOLB问题将导致处理器有不必要的空闲时间和非均匀负载模式。在本文中,我们评估的性能立即模式远程3 d图形系统是影响HOLB以及无序执行如何提高性能。




您需要先支付 30元 才能查看全部内容!立即支付