|
|
|
|
|
|
|
|
Non-photorealistic applications of computer graphics date from the days of lineprinter output. Surveyors, cartographers, geographers, geologists, urban planners, meteorologists and mapping agencies were among the pioneers seeking to exploit advances in computer graphics. For example, Coles (1974, p 69) noted that by the end of 1971, after some seven years of R & D, the Ordnance Survey of Great Britain was aware of the need for pilot production schemes for evaluation and further development. Users were also quick to take advantage of developments in 3D computer graphics from the early 1970s. Given the convenience of photorealistic visualisation of digital elevation models, there was no longer much interest in the teaching and deployment of the skilled manual art of landscape drawing. However, Sasada (1987) and Saito and Takahashi (1990) demonstrated the benefits of what the former called "the realism of drawing". Recently, now that the challenge of photorealism has been largely fulfilled, there has been growing interest in the development of hardware and software solutions for Non-Photorealistic Animation and Rendering (NPAR). Gooch and Gooch (online; 2001); Reynolds (online), and Strothotte and Schlechtweg (2002) provide overviews and pointers to the ever growing resources. Other recent texts, such as Akenine-Möller and Haines (2002) include chapters on NPAR. Much of the research effort has been focused on game development, cartoon animation, visualisation of CAD models, simulation of various artistic styles and sketch-driven construction of spatial models. Both line drawings and cartoon (or toon) shading have required the definition of new terms, such as border edges, creases and silhouettes. Developers of rendering algorithms are among the main contributors of such NPAR terminology and their definitions. The CISRG has pioneered algorithms for sketching terrain since the 1990s. This research has drawn on pre-existing ideas in not only cartography and geography but also on the psychology of perception and cognition, vision, image processing and the philosophy of science. Within these disciplines, terms (such as boundary, occluding contour and silhouette) have somewhat different meanings to those being ascribed within NPAR. This multi-disciplinary inheritance of terminology can be confusing and worse still lead to misconceptions of the precise entities under discussion. Also, there seems to be a need for a refinement of the vocabulary. Ramesh Raskar's question, in response to our evaluation of the Polygon Fattening approach to silhouette rendering, provided the catalyst for setting up this discussion thread. I found it difficult to directly answer his question What is the difference between silhouette and occluding contour? since we ourselves were attempting to conceptualise silhouettes differently. As we stated in our evaluation, Formulated Silhouettes were only devised as a proof of concept of a type of imagined or illusory contours which are perceived as the silhouettes of hills, in particular, and perhaps of other forms. As yet, we do not know how to operationalise the rendering of this illusory silhouette in 3D but believe that it is more pertinent than the occluding contours of landforms. Worse still, we do not know what to call the conceptual entity. The term Formulated Silhouette refers to a concocted prototype - it is more like phlogiston than oxygen but, like both of them at their inceptions, it seeks to shift attention to the something other (beyond gross physical matter) which had not been considered within R & D in NPAR. Most dictionary definitions define silhouette as the filled in outline (outer line) of a person or object as seen, after the 18thC French finance minister, Etienne de Silhouette, who cut out the profile outlines of prominent political figures on black paper. Such silhouettes are still produced as an art form (Courtney, online); and since 1986, NASA [online] has developed a hidden line removal program for calligraphic drawings to enhance realism and artistic presentation; this program extracts silhouettes, with and without internal visible detail, from line segments or polygons describing arbitrary solids. Silhouettes with and without in-fill were intensively studied by Gestalt psychologists in figure-ground discrimination, and in the study of visual puzzles (such as ambiguous figures) and the predisposition to see illusory contours as belonging to occluding objects in 2.5D space. Occlusion means shutting off or blocking. An occluding form is one which blocks the view and its outline as seen is referred to as the occluding contour. Marr (1982, p 218) stated "An occluding contour is simply a contour that marks a discontinuity in depth, and it usually corresponds to the silhouette of an object as seen in a two-dimensional projection"; under his Figure 3-60 on p 225, he refers to "silhouette (the circumscribing contour)". In his Figure 4.2 on p 278, he then refers to this outline as the occluding contour. AI researchers in the 1960s and 1970s, such as Waltz (1975, referred to by Marr on p17, had already developed effective and efficient algorithms to infer how objects and their depth relations in line drawings, derived from images of prismatic solids, could be interpreted through junctions indicating occlusion. Thus, the terms silhouette and occluding contour have been used as synonyms. If we move away from an arrangement of disjoint solid objects to surfaces, such as of terrain, the objects of interest in the scene are the constituent landforms. The circumscribing contour shows the limits of data, and includes the horizon. The intervening silhouettes of landforms facilitate their recognition and provide a qualitative indication of their depth relations and interposition in space. The definition of silhouettes in NPAR admits the intervening contours and, optionally (see Akenine-Möller and Haines, 2002) exclude parts of the circumscribing outline. Since the latter corresponds to the original definition of a silhouette, this introduces scope for some confusion. More importantly from the perspective of landscape drawing, the silhouette does not just indicate the limits of occlusion, it also functions as a signifier. Semiology is the general science of signs. Bullock et al (1988, p 769) define semiology as: systems of signification, means by which human beings - individually or in groups - communicate or attempt to communicate by signal ... and other things that qualify . Semiology is a branch of semantics; one meaning of which was established by the Polish logician Alfred Tarsky as the study of the RELATION between linguistic expressions and the objects in the world to which they refer or which is their function to describe (Bullock et al, 1988, p 769). This refers to the deeper meaning of signs and not just their superficial meanings as defined by their use within linguistic expressions as defined by denotations and syntax. Semiology and semantics are thus concerned with both syntax and pragmatics, the study of the dependence of the meaning of linguistic expressions on their users, and on the circumstances in which and the purposes for which they are used. The NPAR community encourages artistic licence in the stylised rendering of silhouettes. But the mathematical definition(s) of silhouettes is based on the concept of physical occlusion and disregards how silhouettes might be perceived. The challenge for photorealism was the formulation of algorithms encapsulating the physics and mathematics for simulation of natural phenomena. The current challenge for NPAR is the accommodation of the pragmatics of signifiers (such as silhouettes) within the semiology of graphics to take account of the psychology of perception and cognition. Perception involves interpretation and creative construction and not just sensation. Like a skilled cook, artists do not adhere rigidly to algorithmic recipes. The art of silhouette cutting seeks to portray the essential character of a person's profile, rather than reproduce the occluding outline per se. Cartographers and artists draw silhouettes to not only furnish qualitative depth relationships but also to portray the form of relief elements. They are aware of Gestalt Laws of Organisation, noted below. Bruce et al (1996, p 98) point out that vision extracts accurately the location and orientation of a wide variety of spatial discontinuities in image properties, where there is a sharp change in colour, texture or motion. The 'reality' of illusory contours is reinforced by recent psychological and physiological findings that point to common pathways with contour neurons processing both 'real' and 'illusory' contours in early vision. They pointed out that the lack of a luminance edge does not make a contour illusory. Similarly, I propose that the generic class of silhouettes belong within this class of so-called illusory contours, which are not necessarily dependent on the presence of physical occlusion in a given view but which would include them when the latter are present. The capacity of the brain to group elements was intensively studied by the Gestalt school who formulated various Laws of Organisation. Similarity (e.g. caused by texture changes) can override proximity and generate discontinuities and quite dissimilar objects may be seen as forming a unit because of proximity and good continuation (see Figure 6.11 in Bruce et al, p 108). Good continuation, moreover, occurs in space and time - the latter dimension is often disregarded (perhaps inadvertantly) in definitions of silhouettes. As Palmer (1981, p 11) noted, etymology (discovery of earlier meanings) for its own sake is of little importance, even if it has curiosity value. Natural language accommodates polysemic words whose precise meanings are defined by their use in the language within some specific context. However, the different usages of the term silhouette is confusing and seems unnecessary given its use as a synonym for outline and occluding contour. It would be helpful if we could distinguish between the different meanings of silhouettes and related terms more explicitly through an agreed vocabulary. Like most other multidisciplinary researchers, I am not an expert in all the diverse fields of philosophy of science, neurophysiology, psychology, cartography and linguistics, never mind computer graphics and its various applications. NPAR can become the meeting ground for cross-fertilisation of ideas leading to the birth of new ones. This discussion forum for comparing the concepts underpinning silhouettes and related terminology was instigated to give everyone, especially those who are unable to attend specialist conferences, an opportunity to throw their bit into the melting pot. It would be helpful if this leads to a consistent usage of terms. Even if it does not, there would at least be much greater awareness of usages and a better understanding of nuances in meaning. I do not intend to edit the contributions of others and would prefer to provide a link to input located on authors' sites which would establish their background. If this is not possible, I will post contributions on our website. I hope that a small group of people will volunteer to sift through contributions (including mine and theirs) to propose a framework of terms and their meanings. I look forward to receiving your contributions. Please click here for guidance notes. mahes |
|
References (under construction) Akenine-Möller,
T and Haines E (2002) Real-Time Rendering, A.K.
Peters Ltd., 2nd edition. Bruce, V, Green P R and Georgenson, M A (1996) Visual Perception: physiology, psychology and ecology. Psychology Press, Hove, UK, 3rd edition. Bullock A, Stallybrass G and Trombley, S (ed, 1988) The Fontana Dictionary of Modern Thought, Fontana Press, London, 2nd edition. Coles, D J (1974) Ordnance Survey Digital Mapping, in Automation in Cartography, Proc of Annual Symp of British Cartographic Society (Southampton UK, Sep 1973), The British Cartographic Society, 69 - 85. Courtney K [online] Silhouettes http://pages.prodigy.com/kate/silhouet.htm [last accessed August 2002] Gooch, B. and Gooch, A. (2001) Non-photorealistic rendering Imprint Natick, Ma. : A.K.Peters Gooch, B. and Gooch, A. [online] Non-Photorealistic Rendering http://www.cs.utah.edu/npr/ [last accessed August 2002] Hoffman D D and Richards W A (1982) Parts for recognition, Cognition 18, 65 - 96. Marr, D (1982) Vision, W H Freeman, New York. NASA
[online] Hidden
Line Computer Code with Generalized Silhouette Solution
http://www.openchannelfoundation.org/projects/SILHOUETTE/ [last
accessed August 2002] Palmer, F R (1981) Semantics, 2nd ed., Cambridge University Press. Reynolds, C. [online] Stylized Depiction in Computer Graphics : Non-Photorealistic, Painterly and 'Toon Rendering, http://www.red3d.com/cwr/npr/ [last accessed August 2002] Saito T and Takahashi T (1990) Comprehensible Rendering of 3-D Shapes, Computer Graphics, 24(4), 197 - 206 Sasada T T (1987) Drawing natural scenery by computer graphics, CAD, 19(4), 212 - 218 Strothotte, T and Schlechtweg, S (2002) Non-Photorealistic Computer Graphics: Modeling, Rendering, and Animation, Morgan Kauffman, San Francisco. Waltz, D (1975) Understanding line drawings of scenes with shadows, In The Psychology of Computer Vision, P H Winston, ed, McGraw-Hill, New York, 19 - 91. Whelan J C and Visvalingam M (2002) [online] Formulated Silhouettes for Sketching Terrain http://www2.dcs.hull.ac.uk/CISRG/projects/whelan-1/index.htm |
Cartographic Information Systems Research Group, University of Hull
