- Email: [email protected]
Theory of computer graphics and CAD addressed
Conferences and exhibitions are just around the corner to tempt everyone. There were over 7500 people at this year's DAC, both exhibiting and attending (and laughing and talking to friends and generally enjoying themselves). The magnitude and scope of the DAC keeps it the premier event in design automation year after
year, and this year was another jewel in the crown. Larry O'Neill and his committee can be justifiably proud of DAC 24. The 25th DAC, the silver jubilee, will be held at the Anaheim Convention Center, Anaheim, California, June 1215, 1988. The conference proceedings
for DAC 24 are available from the IEEE Computer Society, 10662 Los Vaqueros Circle, Los Alimitos, CA 90720, USA, Tel (714) 821-8380, as Order Number 781 at $94 the copy list price (members $47). Manfred A Ward
Theory of computer graphics and CAD addressed NA TO International Advanced Study Institute: Theoretical Foundations of Computer Graphics and CAD (4-17 July 1987) II Ciocco Castelvecchio, Tuscaoy, Italy This is a brief report of the II Ciocco Advanced Study Institute, sponsored by NATO and DEC as a follow up to 'Fundamental Algorithms for Computer Graphics' (Springer Verlag Series F: 'Computer and Systems Sciences' Vol 17 staged at llkley in the Spring of 1985). The full conference proceedings of this ASI too will be available from Springer Verlag shortly. II Ciocco is a magnificent Italian conference centre, based on a rambling hotel (mainly modern, though with a 400 year old core) set high in the Tuscan hills above Barga. The conference facilities, managed by the genial 'Bruno' are excellent except for the terraced bar which is infested (along with adjacent bedrooms) from evening to the early hours by an excruciating piano player singing out of tune pop songs from the early fifties along with a metronomic accompaniment of computer synthesized tub thumping, all electronically amplified into kilowatt loudspeakers. The catering under the supervision of 'Alberto', an Italian trained in Glasgow (of all places) was superb, though I for one tended to tire of the endless variations of freshly baked pasta and non-stop local chianti. (I put my dizziness down to the hot weather for, having spent 109 lira on a new helicopter, the hotel could not afford to run its air conditioning.) All 120 speakers and delegates were required, on arrival, to sign an oath of allegiance to the spirit of the Institute, requiring us to attend all of the conference programme as organized by
volume 19 number 10 december 1987
our managing director, Rae Earnshaw of Leeds, a stern believer in the ethic that every available minute should be filled with sixty seconds worth of lecture listened to, with formal sessions from 9.00 in the morning until 10.00 at night (and often later), though mercifully even Rae could not persuade the Italians to settle for a quick lunch, so siesta time, 2.00 to 3.00 p.m. was left 'free' for work. (Personally, I find that the first few lectures always trigger exciting debate with other enthusiasts, and the back two rows of the main lecture hall were grabbed by those listening with one ear while beavering away with pen and calculator. The odd blank sides from our rupturing wad of preprints were soon at something of a premium.) As the title suggests the main themes of the conference were concerned with topics in coml~uter graphics and computer-aided design. However, the computer graphics presentations could be further classified under computational geometry, image processing, data structures for graphic representation and formal methods for the validation of graphics based algorithms: As there were 15 invited lecturers, 11 contributing lecturers and 29 delegate papers actually presented we are unable to refer to them all, but we pick up just a few highlights in the remainder of this report. The sessions began with a presentation by Dr Mark Overmans from the University of Utrecht who gave an overview of existing algorithms for sorting and searching of multidimensional partsets, and the determination of intersections between rectangles and lines segments which constitutes an important step in hidden line removal algorithms. Professor Robin Forrest of the
University of East Anglia emphasized the need to incorporate software engineering techniques in the construction of systems for computational geometry. He also stressed that we should develop specialized languages and hardware and that to do this effectively we need a sound theory for the systems issues as opposed to the theory of algorithms in isolation. As one of the first to coin the term 'computational geometry', Professor Forrest pointed out that all geometric problems tackled through a computer algorithmic approach were necessarily of a discrete nature and that the term 'continuous' currently favoured by some practitioners had little meaning in this context. Continuing in the computational geometry field, Professor David Dobkin of Princeton University identified five key problems 'given a set of points' in the geometry of the plane, and the corresponding development of relevant solutions within the past decade. The problems were identified as • the determinations of the convex hull of a point set • the determination of nearest neighbouts • the subdivision of the plane into specified regions • efficient searching and sorting algorithms for a set of points in n dimensional Euclidean space • the problem of determining the intersection points for intersecting segments Professor Dobkin then outlined a number of algorithms and their development aimed at providing general solutions to these key problems. It subsequently emerged that whilst a considerable number of algorithms of
569
Conferences and exhibitions varying complexity existed for providing suitable solutions to the problems there were many data sets for which a generalized solution could not be found, due largely to the pathological distribution of the data. Consideration of these and related problems have caused the field to embrace more mathematicians in an attempt to obtain more general theoretical solutions to the probler~. As yet therefore there is still a great deal to be achieved in this rapidly expanding area. At this point in the proceedings Dr Tucker of the Centre for Theoretical Computer Science presented a paper dealing with formal methods for the specification and verification of line drawing algorithms and tracings. The approach adopted was to specify the design of an algorithm as a series of draft designs with each design being defined in terms of the formal specification of the logical behaviour of the algorithm, together with a set of proofs and algorithmic tests. The chosen test case was the well known Bresenham line drawing algorithm. It soon became clear, however, that to formally establish the general validity of the algorithm, which can be written as one page of text, required something of the order of thirty pages of text excluding tests. In the debate which followed, one of us (Professor M Pitteway) likened the work to the situation created by the mathematician Bertrand Russell who in the process of attempting to prove that one plus one equals two filled enough sheets of paper to decorate his study and apparently still bad some way to go before completion of the work. For many of us insight is more important than the proof. Improvement in the quality and efficiency of line drawing and image representation algorithms was discussed by a number of participants. In particular, Dr lack Bresenham, well known for his pioneering work on line drawing algorithms, gave an interesting paper on the selection and customization of line drawing algorithms to cope with anomalies such as perturbation effects of clipping, problems of retracing paths and the determination of common end points for lines whose end coordinates differ by only a fractional amount. He went on to illustrate the similarities between the problems of
570
pel-level rastering and the problems of line segment intersection algorithms in geometric modelling. For example to the question 'do two line segments intersect?' the answer must be more than just a simple 'yes' or 'no'. In concluding Dr Bresenham advised that any algorithm aimed at line drawing should be fully understood in terms of its implicit assumptions and quirks, and that it would be for the user to modify the algorithm for a specific purpose. Professor Mike Pitteway, Brunel University, described an integer-based algorithm for producing conic arcs at high resolution and which also reduced the staircasing effect (jaggies) which occur in the drawing of lines by raster displays. The algorithm exploits the grey-scale capability offered by many modern display devices to achieve this softening effect. Professor Pitteway concluded by identifying a number of outstanding problems in the development and implementation of the algorithm. For example do methods such as this, which combine grey-scale manipulation to achieve an aesthetic improvement in image quality, lead to increased operator fatigue compared with the use of higher resolution displays? Also, how many grey-scale levels are necessary to achieve the optimal result? Professor Pitteway said that human factor research could provide an important insight in this field. In this same area Professor Pavlidis of the State University of New York presented a method using the formalism of signal processing and random sampling techniques together with the subsequent use of low pass filtering to improve the quality of the displayed image. Data structure methods for the representation of images such as quadtrees and octrees were described in some depth by Professor Samet of the Computer Science Department, University of Maryland. Representation of information via these hierarchical structures enabled the rapid access of components of the information when required. He emphasized their increasing importance in the areas of computer graphics, computer-aided design, robotics, computer vision and cartography. He explained that their implementation is based upon the principle of recursive decomposition which is
very similar to the methods often referred to as divide and conquer. However, despite the advantages to decomposing image data into quadtrees representation Professor Samet explained that the disadvantage of such methods is that their space requirements are dependent on the position of the origin. He added, however, that for complicated images the optimal positioning of the origin will usually lead to little improvement in space requirements. There were many other significant and major contributions at both 'professional development' and at 'state-of-the-art' levels, as in the following, for example. Umberto Cugini on 'The role of different levels of modelling in CAD systems', Rae Earnshaw on 'New mathematics for computer graphics', James Foley on 'Models and tools for the designers of user-computer interfaces', Henry Fuchs on 'Hardware and VLSI aspects of CAD', Leo J Guibas on 'Ruler, compass and computer: the design and analysis of geometric algorithms', R John Lansdown on 'Graphics, design and artificial intelligence', Wolfgang Strasser on 'Constructing 3D objects from 2D information', Paul ten Hagen on 'A model for geographical interaction', Varol Akman on 'Geometry and graphics applied to robotics', David Avis on 'Algorithms for high dimensional stabbing problems', David A Duce on 'Formal specification of graphics software', John M Duncan on 'Topology, geometry, and graphics for ship CAD', Alan Middleditch on 'The representation and manipulation of convex polygons', Lyle B Ramshaw on 'B6ziers and B-splines as multiaffine maps', Alfred Schmitt on 'Ray tracing algorithms - theory and practice', Micha Sharir on 'Davenport-Schinzel sequences and their geometric applications', G T Toussaint on 'Some collision avoidance problems in the plane' and Y Y Zeevi on 'Computer image generation using localised elementary functional matched to human vision'. Just like llkley, II Ciocco is destined to become a time rather than a place in computer graphics.
M L V Pitteway and A N Barrett (Computer Science Department, Brunel University, Uxbridge, UK)
computer-aided design
year, and this year was another jewel in the crown. Larry O'Neill and his committee can be justifiably proud of DAC 24. The 25th DAC, the silver jubilee, will be held at the Anaheim Convention Center, Anaheim, California, June 1215, 1988. The conference proceedings
for DAC 24 are available from the IEEE Computer Society, 10662 Los Vaqueros Circle, Los Alimitos, CA 90720, USA, Tel (714) 821-8380, as Order Number 781 at $94 the copy list price (members $47). Manfred A Ward
Theory of computer graphics and CAD addressed NA TO International Advanced Study Institute: Theoretical Foundations of Computer Graphics and CAD (4-17 July 1987) II Ciocco Castelvecchio, Tuscaoy, Italy This is a brief report of the II Ciocco Advanced Study Institute, sponsored by NATO and DEC as a follow up to 'Fundamental Algorithms for Computer Graphics' (Springer Verlag Series F: 'Computer and Systems Sciences' Vol 17 staged at llkley in the Spring of 1985). The full conference proceedings of this ASI too will be available from Springer Verlag shortly. II Ciocco is a magnificent Italian conference centre, based on a rambling hotel (mainly modern, though with a 400 year old core) set high in the Tuscan hills above Barga. The conference facilities, managed by the genial 'Bruno' are excellent except for the terraced bar which is infested (along with adjacent bedrooms) from evening to the early hours by an excruciating piano player singing out of tune pop songs from the early fifties along with a metronomic accompaniment of computer synthesized tub thumping, all electronically amplified into kilowatt loudspeakers. The catering under the supervision of 'Alberto', an Italian trained in Glasgow (of all places) was superb, though I for one tended to tire of the endless variations of freshly baked pasta and non-stop local chianti. (I put my dizziness down to the hot weather for, having spent 109 lira on a new helicopter, the hotel could not afford to run its air conditioning.) All 120 speakers and delegates were required, on arrival, to sign an oath of allegiance to the spirit of the Institute, requiring us to attend all of the conference programme as organized by
volume 19 number 10 december 1987
our managing director, Rae Earnshaw of Leeds, a stern believer in the ethic that every available minute should be filled with sixty seconds worth of lecture listened to, with formal sessions from 9.00 in the morning until 10.00 at night (and often later), though mercifully even Rae could not persuade the Italians to settle for a quick lunch, so siesta time, 2.00 to 3.00 p.m. was left 'free' for work. (Personally, I find that the first few lectures always trigger exciting debate with other enthusiasts, and the back two rows of the main lecture hall were grabbed by those listening with one ear while beavering away with pen and calculator. The odd blank sides from our rupturing wad of preprints were soon at something of a premium.) As the title suggests the main themes of the conference were concerned with topics in coml~uter graphics and computer-aided design. However, the computer graphics presentations could be further classified under computational geometry, image processing, data structures for graphic representation and formal methods for the validation of graphics based algorithms: As there were 15 invited lecturers, 11 contributing lecturers and 29 delegate papers actually presented we are unable to refer to them all, but we pick up just a few highlights in the remainder of this report. The sessions began with a presentation by Dr Mark Overmans from the University of Utrecht who gave an overview of existing algorithms for sorting and searching of multidimensional partsets, and the determination of intersections between rectangles and lines segments which constitutes an important step in hidden line removal algorithms. Professor Robin Forrest of the
University of East Anglia emphasized the need to incorporate software engineering techniques in the construction of systems for computational geometry. He also stressed that we should develop specialized languages and hardware and that to do this effectively we need a sound theory for the systems issues as opposed to the theory of algorithms in isolation. As one of the first to coin the term 'computational geometry', Professor Forrest pointed out that all geometric problems tackled through a computer algorithmic approach were necessarily of a discrete nature and that the term 'continuous' currently favoured by some practitioners had little meaning in this context. Continuing in the computational geometry field, Professor David Dobkin of Princeton University identified five key problems 'given a set of points' in the geometry of the plane, and the corresponding development of relevant solutions within the past decade. The problems were identified as • the determinations of the convex hull of a point set • the determination of nearest neighbouts • the subdivision of the plane into specified regions • efficient searching and sorting algorithms for a set of points in n dimensional Euclidean space • the problem of determining the intersection points for intersecting segments Professor Dobkin then outlined a number of algorithms and their development aimed at providing general solutions to these key problems. It subsequently emerged that whilst a considerable number of algorithms of
569
Conferences and exhibitions varying complexity existed for providing suitable solutions to the problems there were many data sets for which a generalized solution could not be found, due largely to the pathological distribution of the data. Consideration of these and related problems have caused the field to embrace more mathematicians in an attempt to obtain more general theoretical solutions to the probler~. As yet therefore there is still a great deal to be achieved in this rapidly expanding area. At this point in the proceedings Dr Tucker of the Centre for Theoretical Computer Science presented a paper dealing with formal methods for the specification and verification of line drawing algorithms and tracings. The approach adopted was to specify the design of an algorithm as a series of draft designs with each design being defined in terms of the formal specification of the logical behaviour of the algorithm, together with a set of proofs and algorithmic tests. The chosen test case was the well known Bresenham line drawing algorithm. It soon became clear, however, that to formally establish the general validity of the algorithm, which can be written as one page of text, required something of the order of thirty pages of text excluding tests. In the debate which followed, one of us (Professor M Pitteway) likened the work to the situation created by the mathematician Bertrand Russell who in the process of attempting to prove that one plus one equals two filled enough sheets of paper to decorate his study and apparently still bad some way to go before completion of the work. For many of us insight is more important than the proof. Improvement in the quality and efficiency of line drawing and image representation algorithms was discussed by a number of participants. In particular, Dr lack Bresenham, well known for his pioneering work on line drawing algorithms, gave an interesting paper on the selection and customization of line drawing algorithms to cope with anomalies such as perturbation effects of clipping, problems of retracing paths and the determination of common end points for lines whose end coordinates differ by only a fractional amount. He went on to illustrate the similarities between the problems of
570
pel-level rastering and the problems of line segment intersection algorithms in geometric modelling. For example to the question 'do two line segments intersect?' the answer must be more than just a simple 'yes' or 'no'. In concluding Dr Bresenham advised that any algorithm aimed at line drawing should be fully understood in terms of its implicit assumptions and quirks, and that it would be for the user to modify the algorithm for a specific purpose. Professor Mike Pitteway, Brunel University, described an integer-based algorithm for producing conic arcs at high resolution and which also reduced the staircasing effect (jaggies) which occur in the drawing of lines by raster displays. The algorithm exploits the grey-scale capability offered by many modern display devices to achieve this softening effect. Professor Pitteway concluded by identifying a number of outstanding problems in the development and implementation of the algorithm. For example do methods such as this, which combine grey-scale manipulation to achieve an aesthetic improvement in image quality, lead to increased operator fatigue compared with the use of higher resolution displays? Also, how many grey-scale levels are necessary to achieve the optimal result? Professor Pitteway said that human factor research could provide an important insight in this field. In this same area Professor Pavlidis of the State University of New York presented a method using the formalism of signal processing and random sampling techniques together with the subsequent use of low pass filtering to improve the quality of the displayed image. Data structure methods for the representation of images such as quadtrees and octrees were described in some depth by Professor Samet of the Computer Science Department, University of Maryland. Representation of information via these hierarchical structures enabled the rapid access of components of the information when required. He emphasized their increasing importance in the areas of computer graphics, computer-aided design, robotics, computer vision and cartography. He explained that their implementation is based upon the principle of recursive decomposition which is
very similar to the methods often referred to as divide and conquer. However, despite the advantages to decomposing image data into quadtrees representation Professor Samet explained that the disadvantage of such methods is that their space requirements are dependent on the position of the origin. He added, however, that for complicated images the optimal positioning of the origin will usually lead to little improvement in space requirements. There were many other significant and major contributions at both 'professional development' and at 'state-of-the-art' levels, as in the following, for example. Umberto Cugini on 'The role of different levels of modelling in CAD systems', Rae Earnshaw on 'New mathematics for computer graphics', James Foley on 'Models and tools for the designers of user-computer interfaces', Henry Fuchs on 'Hardware and VLSI aspects of CAD', Leo J Guibas on 'Ruler, compass and computer: the design and analysis of geometric algorithms', R John Lansdown on 'Graphics, design and artificial intelligence', Wolfgang Strasser on 'Constructing 3D objects from 2D information', Paul ten Hagen on 'A model for geographical interaction', Varol Akman on 'Geometry and graphics applied to robotics', David Avis on 'Algorithms for high dimensional stabbing problems', David A Duce on 'Formal specification of graphics software', John M Duncan on 'Topology, geometry, and graphics for ship CAD', Alan Middleditch on 'The representation and manipulation of convex polygons', Lyle B Ramshaw on 'B6ziers and B-splines as multiaffine maps', Alfred Schmitt on 'Ray tracing algorithms - theory and practice', Micha Sharir on 'Davenport-Schinzel sequences and their geometric applications', G T Toussaint on 'Some collision avoidance problems in the plane' and Y Y Zeevi on 'Computer image generation using localised elementary functional matched to human vision'. Just like llkley, II Ciocco is destined to become a time rather than a place in computer graphics.
M L V Pitteway and A N Barrett (Computer Science Department, Brunel University, Uxbridge, UK)
computer-aided design