Logo To EPFL

Accepted RIDT'98 Papers and Posters



Title: Using Typography in Document Image Analysis
Authors: Frederic Bapst and Rolf Ingold, IIUF - University of Fribourg (Switzerland)
Pages: ~12 pages
Summary: Even if font usage plays an important role in Document Image Analysis (DIA), recognition systems generally take the concept of font management in a weaker sense than in the production cycle. With the point of view of the document recognition community, we show how typographic information (characters bitmap, metrics, etc.) can improve existing analysis methods. After a brief survey of font recognition issues, we present the advantages of a font software support in the design of recognition systems. Concrete algorithms are proposed in the sub-topics of a posteriori font recognition, monofont Optical Character Recognition (OCR), and word segmentation. The reported experiments and results indicate that there are still substantial benefits to expect from typography-aware analyzers.


Title: Font Decoration by Automatic Mesh Fitting
Authors: Cameron Browne, Canon Information Systems Research, Australia
Pages: ~16 pages
Summary: Glorious Fonts, a new graphically-based font type, allow interesting effects to be applied to existing character sets. The difficulties of automating this process are examined in the context of applying Celtic knotwork designs to plain text characters. An analysis of knotwork construction techniques shows how complete designs may be described by an expanded tile set. This paper also presents an algorithm for automatically fitting a flexible mesh to arbitrary character outlines, based on a semi-regular distribution of internal points and their Delaunay triangulation. This mesh is of particular interest as it is composed of well-formed quadrilateral elements, with few supporting triangular elements, upon which the knotwork design may be realised.

Connare, invited paper

Title: Designing and Using Scalable Type for Display as Graphic Elements on Digital Devices
Authors: Vincent Connare, Microsoft, Redmond, USA
Pages: ~10 pages
Summary: Today's typographic designs are displayed on many different output devices. The type designer can not make a single letterform and expect that single representation to remain faithful to the original artwork. Intelligent font files can provide a type designer the mechanism to tailor the image for multiple screen based devices.


Title: Image Mosaics
Authors: Adam Finkelstein and Marisa Range, Computer Science Department, Princeton University
Pages: ~14 pages
Summary: We describe a process for creating an image mosaic--a collection of small images arranged in such a way that when they are seen together from a distance they suggest a larger image. To visually suggest the larger form, the small images are arranged to match a large picture as much as possible, and then their colors are adjusted to better suggest the overall form. Arrangement of the small images may either be manual or automatic. Adjustment of the colors in the small image to further suggest the larger picture is fully automatic and employs a new color correction scheme that generalizes traditional halftoning.


Title: Producing the Skeleton of a Character
Authors: Jakob Gonczarowski
Pages: ~12 pages
Summary: We present an algorithm that produces a skeleton for an outline character. The algorithm precisely finds the strokes in the skeleton by using the centers of the curves in the character, by scanning the character in prescribed scanline directions, filling the gaps in the scanning result, and deleting unwanted curves/strokes, Finally, the holes at points of topological change, which are left empty by the scanning process, are filled by a heuristic method. The proposed algorithm is demonstrated with several characters, showing also the limitation of the algorithm at its last step.


Title: Simplification of the Arabic Script: Three Different Approaches and their Implementations
Authors: Yannis Haralambous
Pages: ~18 pages
Summary: In 1945, the Cairo Academy for the Arabic Language opened a contest to find the best project for simplification of the Arabic writing system. They received about 200 replies. We have chosen three of these projects (a preliminary project by an Academy subcommitee, and projects by Ahmed Lakhdar-Ghazal and Yahya Boutemene) and have implemented them via the ? typesetting system. In this paper we describe and discuss these projects and their implementations. A text sample is presented both in simplified and in regular form. Showing new aspects of the Arabic script, these systems can be useful for Arabic typesetting, as well as in providing new directions for Arabic type design.


Title: A window-based Method For Automatic Typographic Parameter Extraction
Authors: Jacky Herz, Changyuan Hu, Jakob Gonczarowski, Roger D. Hersch
Pages: ~8 pages
Summary: The synthesis of existing fonts with characters represented by parametrized structure elements requires determining a set of font-specific global and local parameters. Parameters comprise for example, the widths of vertical stems, horizontal bars and curved elements, the spacing between vertical stems, the relative position of the junction between arches and vertical bars and serif measures. These parameters need to be extracted from existing outlines fonts. This paper presents a window-based method for locating within existing outline characters the position of character features from which parameters can be measured. The method is based on the match between outline characters and their corresponding virtual skeletons.


Title: A Unified Table Approach for Efficient Typographic Rendering of Text
Authors: ByungHoon Kang, University of Maryland, USA
Pages: ~12 pages
Summary: As digital documents become widely available, they require efficient text rendering techniques for their aesthetic looks. Rendering techniques are performed on various typographic features. However, in conventional rendering algorithms each typographic feature entails its own data structure (table), and thus the whole rendering requires individual process for each of the features, which causes expensive and redundant computation. The problem is even more critical in a large character set such as Unicode. This paper presents a new optimized text rendering algorithm which is based on a generic unified table. The unified table incorporates all data structures for typographic features, and thus enables rendering to take one-step process. In addition to such typographic features, the unified table can have consequences in text alignment. Intermediate results of computation for text alignment can be naturally incorporated (cached) into the unified table and repeatedly used without recomputing. The new algorithm results in orders of magnitude improvement in text rendering and is expected to enhance current document viewers (e.g., Web browsers) significantly.


Title: A Beta-Velocity Model for Simulating Handwritten Korean Scripts
Authors: Do-Hoon Lee, Miryang National University (Korea)
Pages: ~16 pages
Summary: The shape of oriental languages is different from that of western languages based on alphabet and the writing method is also different. So western handwriting model can not be directly applied to the oriental languages although many handwriting models have proposed for alphabet based languages. In this paper we analysis the Korean handwritten scripts, and parameterize its characteristics. And the parameters are used to generate various handwritten Korean scripts by our proposed model, Beta-Velocity model, which is generalized model of Delta LogNormal model proposed by Plamondon. In order to generate cursive strokes which are consisted of Korean character, we adopt a Beta-distribution curve. The Beta distribution curve basis enables us to make asymmetric velocity model, which was not possible in the previous Delta LogNormal model. Combining these features, we could synthesize different handwriting Korean scripts by controlling parameters. Proposed model is useful to generate the large data for on-line handwritten recognition system and can be supported various and graceful fonts to the commercial word-processor.


Title: Fundamentals of 3D Halftoning
Authors: Qun Lou, Univ. of Zurich, Switzeland
Pages: ~14 pages
Summary: The demands of 3D halftoning technologies stem from rapid prototyping where the reconstruction of volumetric objects of variable density is of broad interest. The implementation of 3D halftoning algorithms can be subdivided into two steps. The first step consists of the extension of 2D halftoning techniques while the second step is to re-adjust the voxel distribution to meet special boundary conditions coming from additive fabrication. Elaborate integration of these two steps will result in high resoluted objects which cannot only mimic the original input density variation but can also be practically built on a stereolithography machine. Such objects are crucial for cartilage and soft-tissues approximation for surgical operation rehearsals and will improve general expression abilities of the parts. In this paper the challenging tasks in 3D halftoning are briefly outlined. The emphasis will be placed on the image processing part rather than additive fabrication dependent part. 3D adaptations of orderd-dither and error diffusion are devised.


Title: The formulation of parameters for type desing based on calligraphic studies
Authors: S.K. Mohanty, Centre for Development of Advanced Computing, Pune India
Pages: ~14 pages
Summary: There is a great degree of structural diversity and complexity in the existence in Indian scripts. It is a hindrance for type designers to design a vast range of type styles in Indian scripts. In this study different kinds of calligraphic tools and their impressions on the paper have been identified for the purpose of type designing. A number of parameters are formulated based on the calligraphic study for a better understanding of the anatomy of letter forms. This methodology contributes in tackling different intricacies at the design stage for type design, particularly in the context of complex Indian scripts.


Title: The Design and Use of a Multiple-Alphabet Font with Omega
Authors: Yannis Haralambous and John Plaice, University of New South Wales Sydney
Pages: ~14 pages
Summary: The Omega project aims to offer open and flexible means for typesetting different scripts. By working at several different levels, it is possible to offer natural support for different languages and scripts, and strictly respect typographical traditions for esch of them. This is illustrated with a large PostScript font for the commonly used left-to-right non cursive alphabets, called "omlgc" ( Omega Latin-Greek-Cyrillic). This font, which more than covers the Unicode sections pertaining to those alphabets, as well as those of IPA, Armenian, Georgian, and Tifinagh (Berber), id built -virtually- out of smaller glyph banks. The Omega typesetting engine, based on that of TEX, is used to print documents using this font. The characters can be accessed either directly, or through the use of filters, called Omega Translation Processes ( Omega TPs ), which are applied to the input stream.

Ostromoukhov, invited paper

Title: Mathematical Tools for Computer Generated Ornamental Patterns
Authors: Victor Ostromoukhov
Pages: ~30 pages
Summary: This article presents mathematical tools for computer-generated ornamental patterns, with a particular attention payed to Islamic patterns. The article shows how, starting from a photo or a sketch of an ornamental figure, the designer analyzes its structure and produces the analytical representation of the pattern. This analytical representation in turn is used for producing a drawing which is integrated into a computer-generated virtual scene. The mathematical tools for analysis of ornamental patterns consist of a subset of tools usually used in the mathematical theory of tilings such as planar symmetry groups and Cayley diagrams. A simple and intuitive step-by-step guide is provided.


Title: An interface for the interactive design of artistic screens
Authors: N. Rudaz, R. D. Hersch, V. Ostromoukhov
Pages: ~12 pages
Summary: This work presents the concepts and the tools involved in the interactive design of artistic screens. The screen elements are derived from a small set of analytical contours provided by the screen designer. We present the requirements that these contours must satisfy in order to generate consistent screens. Software tools have been developed which provide automatic means for verifying and enforcing these constraints. They include a way of specifying the periodicity of the screen dot and a graphical interface offering a convenient way of specifying and tuning the growth of the screen dot.


Title: An Object-Oriented Model for the Hierarchical Composition of Letterforms in Computer-Aided Typeface Design
Authors: Uwe Schneider, Institut für graphische Datenverarbeitung
Pages: ~16 pages
Summary: The scope of this paper is to present an object-oriented model for the hierarchical composition of letterforms within the typeface design process. The model is stroke-based in order to reflect the process of drawing letterforms most naturally. Style is kept consistent by a constraint mechanism, which becomes very powerful by constraining stroke properties instead of outline control points. Especially, non-linear dependencies in style can be expressed. The presented model is used as the underlying paradigm of DaType, a computer aided typeface design system which is currently under development at the Fraunhofer-IGD in Darmstadt, Germany. The major goals of DaType are to closely resemble traditional typeface design techniques as most likely to be accepted by the designer community, as well as to significantly speed up the design process. The DaType approach is not restricted to a certain class of characters. Especially, it may be used for the design of non-latin characters as well.


Title: Feature-Based Design of Fonts Using Constraints
Authors: Ariel Shamir and Dr. Ari Rappoport, The Hebrew University Jerusalem
Pages: ~16 pages
Summary: Computer-aided font design is still a very tedious and repetitions task. New font models and new functionalities presented by digital media further require a new method for the design of font families and dynamic variable fonts. Current design paradigms are either outline based, offering low level operations, or parametric but non visual through programming. Parametric feature based design helps in creating a design environment for high level parametric font design. However, the approaches suggested so far are too restrictive in their definition of features, and none of them utilises constraint in feature definition and design. We present a new approach for parametric feature based font design. Our approach extends the definition and use of features by defining inexplicit feature parts using constraints. Furthermore, the visual design system ans the use of constraints for preserving the designer's intentions create a more natural environment in which high level parametric behaviors can be defined. Thisnot only simplifies the design of single fonts but also enables the simultaneous design of font families by making use of parametrization.


Title: Metafont in the Rockies: the Colorado typemaking project
Authors: Richard Southall, Leckhampstead (UK)
Pages: ~16 pages
Summary: Not available


Title: Visual TrueType: A Graphical Method for Authoring Font Intelligence
Authors: Beat Stamm, Microsoft Inc.
Pages: ~14 pages
Summary: TrueType fonts can be equipped with custom font intelligence for highest quality rendering down to the smallest type sizes at screen resolutions. However, to do so requires programming in an assembly language, which is outside the typical conceptual range of a typographer. In this paper we introduce a static representation of font intelligence in the form of a data structure we call the glyph structure. This glyph structure is visualized and authored graphically, and serialized automatically into the TrueType assembly language. The result is a professional tool that makes instructing high-quality TrueType fonts available to any typographer, without requiring programming skills.



Title: Text rasterisation from 2D to 3D - an experimental framework using texture mapping
Authors: William Harris and Martin J. Duerst, Institut für Informatik der Universität Zürich
Summary: Current research in 3D systems and environments has produced high-quality results for traditional graphical representations. However, non-decorative text, serving as a substantive information-transfer medium, remains unintegrated. Through the combination of traditional 2D rasterising technologies and the innovation of novel 3D rendering algorithms in an experimental framework, a more flexible approach has been developed to support high-quality text representations in 3D fields of application such as user interface design, VR, and scientific visualisation.

Last modified: Thu Jul 15 09:22:01 CEST 1999