Horyuji: imagemaps, plans and panoramas

Michael Greenhalgh
Australian National University
Osaka 2002

Introduction:
the feasible versus the practicable

  1. This paper presents the point of view of a lecturer wishing to place large quantities of (art historical) multimedia before students;
  2. Effective student use of images certainly needs attractive display technologies to explicate complicated spatial relationships;
  3. The projects should be simple enough for the lecturer to set up and replicate across different buildings or ensembles;
  4. Yawning gulf between feasibility and practicability: just because something can be done under lab/mockup conditions does not mean that it works easily outside the laboratory;
  5. Thus laboratory demonstrations must be backed up by demonstrations of practicability, perhaps including estimates of cost;
  6. Elements of practicability therefore include:
    1. ease of project generation;
    2. ease of use: web browser with suitable plugin (usually over a network); NB applications which do not work in a web browser are arguably ipso facto impracticable;
    3. ease of replication into other similar projects;
    4. ease of updating and, if need be, extension (for example by adding to the number of images or panoramas, or video and sound);
  7. Learning using a computer and web browser need not approximate to a video game: accurate, reliable information recollected in tranquillity is much more valuable than jazzy and quickly outdated screenfulls of graphics. In this assertion (to use an art-historical analogy) I am a Cistercian, preferring the beauty of plain architecture to the grotesques of over-decorated cloisters!
  8. As always, the hardware is a relatively trivial cost in comparison with personnel; so the simpler the technologies, the lower the cost, and the more information will reach the users;

The Multimedia Problem

  1. Large quantity of similar-sized images required for good coverage in order to make panoramas or stereo pairs;
  2. Trial-and-error for panoramas increases the quantity needed;
  3. No results yet seen by me which show convincing VRML or other 3D construction of realistic and detailed space of a quality useful for research or teaching (I am not attacking the use of VRML for the presentation of small objects, to which the mechanics of the technology can be well suited);
  4. No programs yet seen by me which offer "effortless" construction of 3D space from photographs, stereo-video, or whatever: all have to be hand-crafted, specially programmed, or both;
  5. No financial statements yet seen by me convince me that VRML and the like are practicable for the ordinary, under-resourced teacher (even assuming an increase in perceived quality);

Solution: the age-old keep it simple

A classically simple approach will probably look good in ten years time, whereas nothing dates so quickly as graphics (cf. 1995 VRML).
  1. Forget the highly-skilled "flashy" technologies: 3D modelling, VRML, and anything that aims to construct virtual worlds, because
    1. difficult to build, requiring a high level of skill;
    2. slow to build: the higher the level of detail, the greater the time needed;
    3. although effective for small (e.g. turntable-loadable) objects, inaccurate, lacking in verismo and inflexible for deasling with large objects occupying real space;
    4. difficult to update;
  2. Stick to photographs (plenty of them), supplemented by video and plans;
  3. Use imagemaps to link elevations and plans to views;
  4. Use maps/floorplans linked to panoramas;

Flexibility and Ease of ImageMaps

  1. Mapping involves relating one "control" image to another, or to a series:
    1. comparanda give context, using a map or plan;
    2. an elevation can act as control mapped to a series of details;
    3. an aerial photograph can also be used together with elevations;
    4. a panorama by definition widens the context of a building or site;
    5. any image can be mapped to any others, in series, just like a set of Russian dolls;
  2. Why use such techniques rather than VRML?
    1. quick and relatively easy;
    2. accurate, with panoramas scalable;
    3. flexible: plans/elevations can be mapped to any level of detail;
    4. works in any browser (locally or across the web) which knows java;
    5. easy to update: just change/add to the photographs;
    6. easy to boilerplate: set up one plan/elevation to your satisfaction, then multiply and change slightly;
    7. should not look out-of-date (important in projects planned to last) - and try looking at a 1996 piece of VRML to underline this point;
  3. Panoramas?
    1. any size; relatively easy to construct;
    2. batch processing sometimes works;
    3. as the pixellation of digital cameras gets better, so one-shot solutions using fisheye lenses or 360-degree cones will become practicable (e.g. 360-One-VR, or the www.behere.com product); currently, the results are not excellent, because of insufficient resolution;
  4. What about video?
    1. easily linked in to imagemaps, triggered by a mouseclick;
    2. can provide "internal panoramas" in "difficult", such as where tripods are discouraged;
    3. in low light conditions, a good substitute for still photographs, because DVCam CCDs work to much lower light-levels than those of digital still cameras;
    4. Zoom capabilities (over 20x non-digital zoom) allow good closeups of distant architectural features impossible to capture with most digital still cameras;
    5. individual DV frames (albeit at a relatively low resolution) can be duplicated as JPEG images, or a sequence ditto;
    6. stereo video is feasible, but where are the demonstrations that it is practicable?
  5. Big or bigger?
    1. today's digital cameras can routinely produce 4mp images and upwards;
    2. digital camera lenses are getting wider and longer: 24mm and 280mm are now possible; fisheye addon lenses may be used; and Nikon's D1 body (circa US$2,000) takes interchangeable lenses;
    3. panoramas can consequently be of 10mp or more, offering good detail and excellent zoomability;
    4. such monsters could be cut down for use today, and brought into use as display technologies get faster;
    5. good software allows relatively painless stitching of panoramas, and we live in hope that automatic stitching of stereo pairs may be possible one day soon;
    6. As well as relying on a web browser across a network, the (large) image could be bundled into an executable, for download and local execution;
  6. Stereo?
    1. stereo adds verisimilitude in studying architecture and sculpture (even bas-reliefs);
    2. digital stereo pairs are possible using small cameras (e.g. Pentax Optio 430s - 4mp) movable on an aluminium rail - total cost under US$2500, and total weight under 1kg. But given the hit-and-miss nature of the process a large number of images is advisable, and digital is much cheaper and more convenient to use than analogue;
    3. there is a fallback: even if not all your stereo pairs "work", each individual image should still produce valuable information;
    4. masochists may wish to attempt stereo panoramas, because the possibilities of it going wrong are doubled when trying to fit together the output from two cameras;
    5. rich masochists might prefer stereo-video: give some details;

    Examples

    1. A plan of the West Precinct with hotspots keyed to panoramas, and the bpanoramas in their turn keyed to a field-of-view on the plan, is here;
    2. A groundplan of the West Precinct keyed to ordinary photographs is here:
    3. The setup also works with elevations or sections, as in this setup for the Chigi Chapel, SM del Popolo, Rome where moving the mouse over the section to the left displays larger-scale details in the window to the right;
    4. the above techniques can be used, with plans keyed to large stereo images which can be treated as panoramas - as in this example of the gate to the west complex;
    5. Large images (which might be panoramas, or stereo anaglyphs) can be packaged for downloading, using pixscreen (www.pixaround.com) software, as in this image of a deity in the gate of the west precinct (here), or in this panorama;
    6. Stereo viewing over the web is easy using Andreas Petersik's java applet, which allows several viewing methods, including anaglyph and JPS. A demonstration page of .JPS pairs (which Andreas' viewer can mogrify into other formats) is here;


    Email: Michael.Greenhalgh@anu.edu.au
    Website (over 180,000 images): http://rubens.anu.edu.au