Ignatius, Eve & Senay, Hikmet
Data, vocabulary, marks, composition rules and visual perception rules
An conceptual overview of the knowledge which is necessary for scientific data visualization
(This information is partially taken from the text [SEN94], in which Vista, the automated visualization system is described.)
If you have read the information about VISTA, an automated system by the authors, you have something read about marks. Well, marks are the most primitive graphical objects, which constitute the graphics. We will explain the concept of marks later. First we will give an overview of the content of this side.
Ignatius and Senay classified the knowledge, which is necessary for scientific data visualization in five categories:
Under use with these five points, we will describe the individual concepts, declared by Ignatius and Senay. We will picture the marks and the composition rules especially.
- data characteristics,
- visualization vocabulary,
- primitive visualization techniques,/li>
- composition rules, and
- visual perception rules.
- Data characteristics
You can divide scientific data into two groups: the qualitative and quantitative data.
- Qualitative data can be subdivide further-on into nominal and ordinal data types. Nominal data types are assemblages of symbolic names, typically unordered. For example, the names of the discoverer, such as Gallileo, Columbus, Magellan, Marc'o Polo etc., form a nominal data set. Ordinal data types are rank ordered only. The ordering of the data does not reflect the magnitudes of the differences. A typical example of an ordinal data set is the names of the calendar month, January through December.
- Quantitative data is more common than qualitative data in all scientific disciplines. They have concrete values like reals. Quantitative data can be scalar, vector and tensor. A scalar possesses a magnitude, but no directional information other than a sign. A scalar is simply defined as a single number. Vectors have in contrast of scalars, both, direction and magnitude. They require a number of scalar components, equal to the dimensions of a coordinate system. A Vector is generally a unified entity, consisting of several scalar-values. Tensors includes several scalar components that change in a particular way during transformation from one coordinate system to another. The number of components that specify a tensor depends on the dimensionality of the coordinate system and the order of the tensor.
- "Other important data attributes that play a role in selecting visualization primitives include functional dependencies among data variables, spacing between sampling points, cardinality of the data set, upper and lower bounds of values, units of measurement, coordinate system, scale, and continuity of data."
- Visualization vocabulary
"The visualization vocabulary identifies the basic building blocks of scientific data visualization techniques. In data visualization, a mark - any graphical symbol visible on a display medium - is the most primitive building block that can encode some useful information." A mark can be simple or compound. There are four types of simple marks: points, lines, areas, and volumes. A compound mark is a collection of simple marks that forms a single perceptual unit. For example: contour lines, glyphs, flow ribbons, and particles are all compound marks.
Marks have positional, temporal, and retinal properties. These ate generally encoded in the data content. A positional encoding shows the variation of the marks' position. A temporal encoding shows how the marks' properties vary over time. A retinal encoding is any variation of the marks that the retina of the eye can perceive besides position.
"Marks can be further classified by whether they represent single or multiple data variables and single or multiple data points." Single variable(SV) marks are aligned with a singe variable, whereas the multiple variable(MV) marks are aligned with several variables. A single data(SD) mark conveys a single value for a single data point. A multiple data(MD) mark, on the other hand, "shows a range of summary information regarding the local distribution of several data points. This classification is particularly useful when visualizing large, multivariate data sets."
The Figure shows the primitive visualization techniques VISTA supports
- Primitive visualization techniques
"Primitive visualization techniques encodes one dependent and up to four independent variables. In general, each primitive visualization technique falls into one of three categories: positional, temporal, and retinal, depending on the primary mark property it manipulates." Positional techniques can be one-, two- or three-dimensional. The only temporal technique is animation. Retinal techniques correspond to the set of retinal properties of marks.
In the figure you see the primitive visualization techniques VISTA supports. You can take them as an example of the different positional- or retinal- techniques. Some of these techniques seemed to be compositions of others, but the authors of VISTA chose that they are primitive visualization techniques too, rather than terms of others. "While it provides a higher level of abstraction for classifying primitive visualization techniques, this organization also tends to reduce design time for complex, multidimensional visualization techniques."
- Composition rules
Composition rules are necessary for displaying multidimensional data. They define conditions under which Vista can combine a pair of visualization techniques. There are five composition rules with which you can describe a large amount of composite visualization techniques:(These composition rules are used by VISTA)
- Mark composition "merges marks of the component visualization technique by pairing each mark of one technique with a compatible set of marks of the other"
- Composition by superimposition "merges marks of the component visualization techniques by superimposing one mark set onto the other"
- Composition by union "combines marks of a pair of component visualization techniques using set union"
- Composition by transparency "combines a pair of visualization techniques by manipulation the opacity values of marks belonging to either or both visualization techniques."
- Composition by intersection "combines a pair of visualization techniques by first computing their intersection, then superimposing the intersection onto one of the components."
Figure 2 shows the different composition rules. (a) mark composition, (b) composition by superimposition, (c) composition by union, (d) composition by transparency, (d) composition by intersection.
- Rules of visual perception
The designing of a visualization is quite a difficult task. You can change the outfit of an visualization by simply change some variables. However, the aim is to create a sufficient effective visualization. To reach this aim, it is necessary that you have knowledge of principles and rules of visual perception. It is quite useful to know, how the resulting image is perceived.