3D virtual scenes provide rich data visualization capabilities and can be used as a medium for presentation of large and complex multi-dimensional datasets. By presenting data in 3D, one gains three clear advantages, which are of critical importance for data visualization systems: large information capacity, enhanced user cognition, and interactivity. Information capacity is of primary importance for large sets of multidimensional data. Following Tufte [80] “even our language, like our paper, often lacks immediate capacity to communicate a sense of dimensional complexity,” Three-dimensional objects can provide information in the form of shapes, colors, textures, positions, sizes, orientations, and even behavior. 3D space is not limited in size, so the only limiting factor for visualization is user perception.
The second advantage is enhanced user cognition. Spatial metaphor for representing data is closer to the manner in which humans perceive the surrounding world. Contrary to flat 2D graphics, a 3D environment permits a user to change the viewpoint to improve perception and understanding of the observed data. A user trying to discover meaning of objects in 3D space, may rotate or translate the objects, if their meaning is not clear at first sight. Information presented in this way is usually learned faster and more efficiently [4, 81].
The third advantage is interactivity, which does not only mean navigation in the space but also interaction with the content, e.g., moving and rotating objects and selecting objects that are of interest to a user. Such interaction capabilities with 3D content in a 3D scene may be enriched with additional 2D interface elements familiar to users accustomed to windows-based interfaces. Such a dimensionally-rich environment gives higher expression power than 2D graphics.
However, employing 3D virtual scenes to data visualization, besides all its advantages, causes also serious difficulties. The most important difficulties are: occlusions, complexity of navigation, lack of easy-to-use 3D pointing devices, limitations of user perception as well as difficulties with presentation of textual information and submission of user inputs and queries. Therefore, 3D visualization should be applied in a careful way in well-justified cases. 3D interfaces should be built with particular consideration of their usability.
3D virtual scenes have been applied to data visualization in many research and production systems. 3D interfaces can be successfully used for presentation of geometrical or spatial data in such domains as medicine, geography, chemistry, astronomy, and physics. In those domains, the volume, nature, and properties of data are in most cases known prior to visualization. Problems appear, however, when 3D interactive virtual scenes are employed for visualization of abstract data without inherent spatial or temporal aspects. The problem becomes even more complex if the structure of data and data volume change from one dataset to another as, for example in the case of search result visualization. Then, application of the 3D visualization metaphor is difficult. In most approaches to this problem, one universal 3D virtual scene is used to present all the data [19, 56, 72]. If the volume of data to be visualized changes significantly, an attempt to use the same virtual scene structure for all possible datasets in many cases causes improper presentation of data. Moreover, often visualization is performed with an assumption that one 3D object should represent one data item. Such an approach, though adequate for small datasets, leads to visual overloading of interface in case of large sets of data. In most cases, selection of interface suitable for visualization of a particular set of data depending on the properties of the data set is not possible.
In most of the existing systems, user interaction is limited to selection of viewpoints and selection of one of the items for presentation [11]. Manipulation of 3D objects (e.g., moving, rotating, and changing attributes), even if possible, provides no additional visualization functionality [37]. Furthermore, mapping of the dataset properties to visual attributes of 3D environments is arbitrary and cannot be altered by a user [72].
Most importantly, in many existing systems, a specialized software is used, very often platform-specific, which requires installation of additional tools, plug-ins, etc. [85]. In most cases, proposed visualization systems do not conform to the 3D content standards such as VRML or X3D [89].
Therefore, research is needed on how to build useable and efficient 3D visualization interfaces for dynamic and complex datasets using 3D content standards.
