Una primera aproximación a la clasificación

Many techniques exist to provide a portion of information which is part of the Illustration Layer. These techniques are adapted according to the type of information to be presented (e. g., text labels, longer texts, images, etc.), the dimension of the data presentation space (e. g., 2D or 3D), or the manner in which a user can interact with the provided information (e. g., selection of information items). But these techniques also differ in whether or not they modify parts of the Background Layer for data presentation (see Fig. 6.7). In this regard, they can be classified in:

• Integration techniques. Techniques that aim at presenting the information close- by their counterparts without modifying the Background Layer. To this end, they try to exploit empty or insubstantial regions. The occlusion of meaningful content is avoided.

• Spatial separation techniques. Techniques that solely use space beyond that occupied by the Background Layer’s visual content. A clear separation between Background and Illustration Layer can be observed.

• Deformation techniques. Techniques that intentionally modify parts of the Back- ground Layer to enable the visualization of other information. The presented infor- mation can correlate with what was replaced.

Illustration Layer display

e.g., Label placement Preserve Background Layer the Modify Background Layer the Techniques that Integration techniques Spatial separation techniques Deformation techniques e.g., Plane projection e.g., Fisheye view

Figure 6.7: Classification of techniques with which information can be presented.

Integration techniques

According to Kakoulis and Tollis [KT98], the integration of text labels into the visu- alization involves the compliance with three basic rules: (1) overlapping of a label with other labels or other graphical features of the drawing should be avoided, (2) it should be easy to identify for each label the graphical feature of the drawing that is addressed, and (3) among all acceptable positions, each label should be placed in the best possible position.

The occlusion of essential parts of the Background Layer can be avoided by identifying its empty or unimportant regions. Ishak and Feiner [IF04], for example, proposed a technique that employs unimportant regions to visualize other data which is in their par- ticular use case the content of windows underneath the focused. Dual-Use of Image Space [CS02] is another technique that makes use of untextured regions for text display. It even goes one step further since the employed space itself is part of the region to be labeled.

The latter technique is also an example for keeping minimal the distance between label and referenced region. The importance of placing labels directly on and around their regions was pointed out by Li et al. [LPS98]. The authors also discussed the aspect that the space which is available around a region decreases with labels placed close-by. Another dynamic technique that focuses labeling of a neighborhood of objects was proposed by Fekete and Plaisant [FP99].

Metadata presentation in 3D space is associated with further challenges since changing the view in three dimensions is feasible. A prominent example is Voxel-Man ([HPP+95, PHP+01]) which is a system that allows to explore the human body interactively. The system comprises various spatial exploration techniques. But textual metadata is basically provided in context menus and by using a simple “object-line-label” technique. There exist many other systems that use variations of the “object-line-label” technique (e. g., [BWHR99, GKH+99]). Preim et al. [PRS97] proposed a system called Zoom Illustra- tor which places annotation boxes at fixed positions beside the 3D model. For certain text items within those boxes, more detailed information can be requested, which is then displayed by employing a zoom technique. G¨otzelmann et al. [GHS06] improved the technique of placing text labels. To this end, they use “agents” that permanently search for adequate positions nearby the model.

Augmented Reality applications demand for intelligent placement of metadata that cor- relates with what a user sees on the display, as well. In this regard, Bell et al. [BFH01, BFH02] proposed techniques for integrating labels and pop-up windows. Their techniques adjust positioning and size of the metadata so that overlapping with important content is avoided.

Spatial separation techniques

Spatial separation of visualization and metadata basically serves two purposes: interfer- ences with the Background Layer are avoided and, at the same time, a considerable amount of data can be presented. A general approach is to attach a new window to the window that displays Background Layer data. The new window, which is visually separated by its frame, can then be employed for showing Illustration Layer data (e. g., [SS00, JGD02]). However, due to long spatial distances between visualization and metadata, it is often difficult to point out correlations. Thus, only metadata for the focused visualization component is shown in most cases.

Including primitives such as illustration planes arranged in 3D space and close-by the visualization of the Background Layer is another technique. Besides adding self-contained display space (e. g., rondell surfaces in [PRS97]), they can also serve for displaying illustra-

6.3. Presentation of Descriptive Metadata 123

tive views of the visualization (e. g., [KDG99]). In this regard, shadow projections are an example for simplifying the original visualization without context loss so that individual components can easily be highlighted (e. g., [RMC91, HZR+92, RSHS03]).

Deformation techniques

The two problems Leung and Apperley [LA94] see when data is to be presented in confined space are a spatial and an information density problem. A technique that solves those presentation problems only in parts was presented by Bouvin et al. [BZGM02]. With the focus on the layout of 2D Web pages, they create space for annotations by moving aside those items that would reach into the space covered by the annotation. However, moving aside items can also yield information loss, particularly when space is confined.

Distortion-oriented techniques, on the other hand, aim at keeping the context while expanding the space for displaying details. Such techniques have been overviewed, clas- sified, and evaluated in various publications (e. g., [LA94, BGBS02, GF04]). Several ap- proaches that employ distortion-oriented techniques were proposed. Examples are: the Document Lens—a tool for navigating in documents with unknown structure [RM93], ex- ploration tools for large information spaces such as graphs, maps, or 3D structures (e. g., [SB92, CCF95, CCF97]), or an approach that allows for displaying Web pages in their entireties [BLH04].

Magic Lenses can also be regarded as a technique that modifies the Background Layer for information presentation (e. g., [BSP+_{93, LH94, VCWP96]). But in contrast to distortion-}

oriented techniques, they do not necessarily magnify the content underneath. Also, there is typically no smooth transition between revealed information and displaced context, even though there is generally some correlation between them.