• Keine Ergebnisse gefunden

C) Generative computer pictures

5 Case Studies: Using the Data Type »Image«

5.2 Rhetorically Enriched Pictures

(“picked” from a given picture) or a color editor. Similarly, texture can be specified verbally by the texture category (in a partially instantiated texture rectangle description) or by an example area from a given picture. Weighed correlation measures are used for computing similarity between two feature vectors. Of course, the most complex type of request can only be stated verbally by naming the object concepts included in the scene, for example, asking for a picture with “mountain”, “snow”, and “lake”; or, on the most general level, by simply asking for the type of scene – “mountain scene”.

Pragmatic restrictions, in particular concerning user modeling, have not been consid-ered in IRIS. However, the image analysis has been explicitly designed in a relatively simple way so that users can more easily understand the categories used for indexing and are not mislead to ask too much “understanding” from the system. Of course, IRIS does not really understand the pictures it analyses; it is able to deliver a coarse approxi-mation to »picture content« based on a simplified picture syntax. This approxiapproxi-mation is proposed as a simple-to-use semi-semantic specification for a kind of image retrieval close to common-sense picture understanding.

Beside its use in picture archiving in the strict sense, the automatically derived de-scriptions can be integrated as special digital watermarks in pictures that are to be pub-lished in the WorldWideWeb: search engines could use that information to find more easily the pictures a user wants to find, and to block others irrelevant (or prohibited for a certain user group).

Figure 108: Query Menue of the System IRIS

and their particular interests), pictures with articulate style differences induce a specific interpretation – at least to some degree. Textbooks for anatomy, for example, avoid traditionally photographs not only because it is too complicated to erase irritating details: essentially, there is too little stylistic variation available with photography’s extreme naturalism. A long tradition of artful anatomical drawing allows the designer for pictures with a better mixture of representation styles fitting their rhetoric demands.

In this case study, we are basically interested in the influence the variation of the de-gree of naturalism has on the (relative) rhetoric function of a pixeme. This also answers corresponding questions in image generation, for example: Which parts of the geometry in question should be presented in a naturalistic way, and why is the rest to be presented in a more “abstract” manner?

Recall in this context the meaning of the expressions ‘realism’ and ‘naturalism’ intro-duced before: ‘realism’ is the property of a representation of giving the impression of a configuration of spatial objects that is or could be found in the world. ‘Naturalism’ re-fers to the degree of a pictorial representation to which it evokes a visual impression as close as possible to that of the scene depicted. While realism is a binary category, natu-ralism only defines one pole of a continuous scale. Compare the two pictures in Figure 109: both represent a spatial scene in representation styles with a low degree of natural-ism. What is the different meaning they suggest by means of their stylistic difference? It obviously is related to the differentiation between aspects (shown as) already known to the beholder, and new (informative) parts. The former can easily be used as anchor points for the other elements supposedly not yet known to the beholder.

Recall also at this place the sub functions of propositional communication (cf. section 3.4.1): by nomination those aspects of an utterance are meant that refer to something al-ready mutually known (e.g., previously mentioned). They are used to provide the other interlocutors with an anchor point for information that is new – the predication. The given anchor and the new distinction can indeed be conceived of as prime functions of rational communication in general. We therefore conceive of them as important rhetoric functions for rhetorically enriched pictures, as well, over and above the more fundamen-tal function as context builder of pictures in general.

The various styles in pictures with mixed representation are only indirectly associated to pixemes. They relate directly to semantic elements, i.e., elements of the picture con-tent. Let us call those parts of the underlying geometric model the ‘representational elements’ of the picture. The part-whole relations inherent to sortal objects organize these elements.

5.2.1 Descibing Style Parameters

In order to select style parameters depending on the rhetoric functions of nomination and predication for an example system, we first have to provide a formal language describing the style variations at hand. As we cannot deal here with the full range of syntactic and semantic factors that may add to the naturalism of a realistic picture, we restrict ourselves to a reduced and very simplified list of visual components (Table 4):

First, there is color: a representational element of a picture may be ordinarily colored, uncolored or colored in an unnatural manner (e.g., duo-toned). Second, representational elements may show texture: the ordinary distribution, no texture at all, or a wrong texture (e.g., cross-hedged). Third, picture elements have form: as a respect for similarity, this dimension ranges from photo-realistically shaded projection of the full 3-D form through sketch with outlines and inner contours for indicating part-whole relations to the pure outline. Finally, the relative place of the representational element and the configuration are relevant: they may be considered as either the “natural” one or any other. The dimensions are not completely independent. Configuration is closely linked with the referents’ part-whole relations, and controls how the corresponding representational elements of the parts form a representational whole. A representational element without parts has no configuration: if a representational element shows only

Figure 109: Two Contrasting Examples of Rhetorically Enriched Pictures

outlines, all its parts are suppressed – the element becomes “atomic”. In that case, the parameters of color and texture have to be adapted, too.

Any presentation style available for tele-rendering can be attributed such a style de-scription. The “classical” photo-realistic rendering for example corresponds to «natural color, natural texture, shaded projection, natural place, natural configuration», a cop-per plate engraving of an exploded view of a technical device to «uncolored, unnatural texture, inner contours, unnatural place, unnatural configuration». The general idea is, that an interactive system uses such a characterization in its active beholder model in order to determine how a picture is to be generated for a particular user, or in its passive beholder model for evaluating a given picture.

In a picture with mixed styles, each representation element has, obviously, its own stylistic characterization: a tree of style descriptions according to the part-whole rela-tions between the elements has to be considered. Due to the dependencies, the attribu-tion of one value in the hierarchy puts constraints on the other values.

5.2.2 The Heuristics of Predicative Naturalism

The dual questions are then: how do we select the style parameters for a given association between representational elements and basic rhetoric functions? And: how can we reconstruct that association given the style parameter description of the representational elements of a picture? We need principles mapping one association of the hierarchy of representational elements with rhetoric basic functions onto another association of the hierarchy with style parameters, and vice versa.

Some presentational elements serve the communicational purpose of anchoring the place of another, usually more central element. We may assume that the beholder of that picture does not yet know that element well enough, which was the original intention to ask for that picture at all. Form and configuration of the other elements are used nomi-natorically, and they are sufficient for the intended user to be able to establish a context already known for the new information. All other respects of representation – color, tex-ture, etc. – are reduced. In contrast to that, the representational element in focus is given a richer, more naturalistic appearance with more details of form and configuration, and, eventually, some color and texture.

Indeed, a representational element of a picture can take over both rhetoric functions:

some of its visual properties may be nominatoric while others are predicative. This ob-servation depends on the fact that there are no pictorial proper names, only definite de-scriptions. A representational element in an image is able to carry the rhetoric function of nomination merely by means of certain visual property in contrast to the other ele-ments. Other style attributes of that element can simultaneously carry the predicative function. We therefore rather speak in the case of pictures of nominatoric and predica-tive properties of representational elements than of nominatoric or predicapredica-tive represen-tational elements.

Table 4 Color Texture Form Place Configuration

natural natural shaded natural natural

uncolored untextured inner contour unnatural unnatural

unnatural unnatural outline atomic

As a heuristic rule derived from these arguments – a Heuristics of Predicative Natu-ralism for realistic pictures suggests itself: the parts of a spatial scene playing the role of nominatoric anchors in a picture of that scene should appear less naturalistic than the representational elements carrying the predicative properties. Correspondingly, the strange and unexpected parts are to be presented more naturalistically than the common and known.

Given a list of (visual) properties to be communicated by a picture and the list of properties that have been communicated already, the active beholder model of an inter-active system has the task to select for each representational element a presentation style according to the style description. That selection must allow the system an encoding so that a subset of the nominatoric properties of that picture element (if there are any) suf-ficient for identification is included. Furthermore, as many of its predicative properties as possible have to be shown. While the nominatoric subset can be reduced to the mini-mal set satisfying a unique identification, the predicational information should be given redundantly to ensure a proper understanding. In accordance with GRICE’s maxims (cf.

Section 4.4.2.2), the picture is then as informative as possible (with respect to predica-tion) and not more informative than is required (with respect to nominapredica-tion).

For a certain communicative intention, this specification enables a system to choose autonomously a style that matches the situational conditions best. The Heuristics of Predicative Naturalism is to be seen as a strategic rule among others for the generation of mix-styled pictures; its results may be overwritten or modified by other strategic rules or meta-principles (e.g., consistency principles).88

The computational visualist can also use the heuristics for critically evaluating the rhetoric force of the interactive system’s pictures by means of the passive beholder model before actually having them presented. According to the Heuristics of Predicative Naturalism, somebody receiving the picture may assume, at least as a first guess, that the producer of that picture wants him or her to understand the less naturalistic parts as nominatoric, and the more naturalistic ones as predicative. The comics examples quoted from MCCLOUD in section 4.4.1.3 can be interpreted exactly in this sense.

5.2.3 Example Application of the Heuristics

In general, the Principle of Predicative Naturalism is realized as a constraint system propagating restrictions through the hierarchy of representational elements and their visual components until a stable association has been established. The following parameters act as additional constraints:

(a) the order of degree of naturalism between the style values (e.g., the degree of naturalism decreases from ‘natural color’ to ‘uncolored’ to ‘unnatural colors’), (b) the dependencies between the different visual components considered by the style

description (color, texture, form, place, configuration), and

(c) the impact a particular attribution of a rhetoric function to a visual component has on the other components of that representational element.

88 Such a set has been suggested, for example, by RIST [1996, Sect. 5.2.2] (cf. section 4.4.2.1).

Figure 110 presents three gray-scale examples: color and texture parameters are ig-nored here for simplicity.89 Let us assume that the communicative intention given asso-ciates the background (a table top) as nominatoric, and a bunny’s shape and the con-figuration of its (body-) parts as predicative, while the relative position of the bunny with respect to the background objects is also marked as known already to the beholders in question. Then, it is sufficient for the background to be drawn in outline and without parts (“atomic”). For the bunny however, form and configuration parameters should be maximal (i.e., value “natural”). If only the bunny’s configuration is predicative, the form parameter changes to “outline”: the four components are clearly discernable.

The third example is more complicated since the place of an element can only be predicative if the configuration of the complete scene is emphasized, as well. Therefore, using the bunny’s place as predicative property has to be propagated up in the hierarchy of representation elements to the configuration of the scene, and from there down again to the place parameter of the other children, i.e., all the background elements. The inner contours of the tabletop are selected while the configuration of the bunny becomes atomic and only outlines need to be shown. The graphic clearly emphasizes the bunny being almost at the border (in consequence, its danger to fall off is here more evident).