Lapas attēli
PDF
ePub

and almost as easy to use as conven- motivation seems to be related to video left-right allowing a great deal of look tional products. I think that we could conferencing, where it is felt that hav- around, and viewing distance can vary make a substantial improvement to ing all the participants view each other over several meters. understanding the complex world with glasses would be very unnatural. Hamasaki, the symposium director, around us by encouraging the use of The basic idea in this approach is has been working in the area of 3D the appropriate parts of this technol- also not new. An object is photographed

also not new. An object is photographed imaging for almost 30 years and, in fact, ogy. (Subsequent to this symposium, with two cameras corresponding to left will be retiring from his current UniStarks wrote to tell me more about the and right eye. Then images are dis- versity of Tokyo position in March 1992. Goldstar prototype TV, which he claims played on a sequence of narrow vertical In addition to the photographic projis soon to appear as a consumer item. It stripes, left eye image, right eye image, ect mentioned above, his laboratory is has line doubling, edge enhancement, left, right, etc., a corduroy or interdig. filled with other interesting experiments and some other simple tricks to improve itated display. These days flat panel associated with autostereoscopic image quality. Unlike previous line display devices are often used for the imaging. One of the most careful involves doublers, this works on a field-by-field displays. Immediately in front is an array a multiplano lenticular display that interbasis, so it is 3D compatible. Starks of half cylindrical lenses roughly digitates eight views behind a lenticule, claims that in 2D it was nearly the equal matching the pitch of the display. Out using an array of eight solid-state senof HDTV and in 3D it was stunning, in front of all this sits the observer, who sor cameras mounted on a base and especially since it is supposed to be can have an authentic stereo image if viewing through a large Fresnel-type available for just a few hundred dollars he/she is positioned in exactly the right optic into a CRT. Using eight (rather more than a standard set and is 100% place. As mentioned above, little head than two) cameras increases the viewcompatible with all existing video hard- movement is possible and the observer ing zone, allows the observer's location ware and software. Starks feels that the must be seated at exactly the right to be less critical, and provides some effective bandwidth of the monitor is position. Multiple observers can view look around ability. But the image is about 40 MHz.)

this kind of display at the same time, still limited in the number of observers In Japan, work in stereo and 3D although each observer must be cor- that can view it. Hamasaki is hoping to imaging spans the same broad subfields rectly positioned. Also it is possible for increase the size of the system to 20 as in the West except that there is a observers to get a pseudostereoscopic cameras. There is a continuing resoludecided difference in emphasis. (At the image (right image to left eye, and left tion problem because the flat LCD symposium, Starks (3D TV) noted that to right).

requires one pixel red-green-blue (RGB) most of the relevant information is not It is also possible to interdigitate set for each camera. There are also given either as journal or conference more than two images, and this reduces promising experiments on projection papers but rather in patent applica- the observer positioning problem, systems using arrays of LC panels. tions. In his paper he presents over 500 although it also decreases the resolu- Considering the variety of projects, the cited patent references to substantiate tion capability of the display devices, a laboratory is doing very impressive this point. Luckily, the Japanese patent problem with pixel-based displays, less research. Symposium participants felt system is the world's first to be comput- so with fully optical displays. NTT incor- that there were few places in the world erized and researchers can now access porates an "eye-tracker" to follow the

engaged in comparable work; howthis information from anywhere in the observer and interchanges the views ever, one observed that the group needs world.) In Japan, there is much more from the lenses if a pseudostereoscopic at least five times the resources available research than in the United States on view would be observed, but this makes to it in order to make timely progress. the study of stereo and 3D in concert it more difficult to use with more than The Heinrich-Hertz Institute in with high definition TV and somewhat one observer. At the University of Tokyo, Berlin is also doing very advanced work more than in the EC. This is natural as Prof. Hamasaki's group has been work- in the design of lenticular screens and both Japan and the EC have HDTV ing with images based on two Fresnel applications to glassless 3D systems. products that are either on the market lenses and 24 prisms in order to make R. Borner presented a summary of these or nearly so. What is somewhat sur- large color photographs. Care is required projects, which have been going on for prising, though, is that the Japanese to produce these, but in the examples I more than 7 years. They are currently are also much more active in research saw, the stereo effect is exceptionally experimenting with an 18-channel front concerning autostereoscopic and glass- realistic and the resolution is as good projector and hope to increase this to less imaging, and especially research as in normal color television. Viewers 24 channels, and they are also working connected with lenticular sheets. A main can move their head about 10 inches on two rear projection systems, one

[ocr errors][merged small]

with a 640x480 pixel LC overhead I was surprised to hear this criti- wide variety of receiver designs as well projection panel in 8 grayscales and cism, as one of the only basic books on as compatibility with 2D systems could the other with an 18-inch direct view this topic was translated into English be achieved. Tilton's idea is to transmit electroluminescent (EL) display with in 1976 from the Japanese edition (1972) one 2D HDTV signal, and another signal 16 grayscales and 1024x864 pixel res- and is widely known here.

describing depth, which he claims can olution. To the best of my knowledge,

be done with a bandwidth of about the Borner group's work designing Three Dimensional Imaging

one-fourth that of the luminance signal. screens with varying pitch corrections Techniques

However, it was clear that most attendees is the most advanced anywhere.

by Takanori Okoshi

felt that bandwidth was a serious probSeveral of the U.S. participants at Academic Press 1976

lem and that related work on image the symposium were surprised at the (Japanese book: Sanjigen Gazo compression was very important. amount of work going on in Japan in Kogaku, 1972)

Several interesting symposium sesthe area of glassless autostereoscopsis.

sions were devoted to applications of (Except for Benton's holographic work My own view is somewhat different. 3D to other disciplines. These included at MIT, there was not much reported the problem of stereo or 3D imaging is analyzing electron micrographs of from the United States on this topic.) old enough that many fundamental ideas GaP_{1-x} for dislocations, medical All researchers would like to dispense have already been proposed. Some of endoscopy, nondestructive investigawith glasses, but most believe that these may have failed in the past because tion of animals' internal organs, etc. practical systems will require them for the implementation technology was not There was a paper by Sony researchers the remainder of this decade. At the up to the demands placed upon it. But on an almost commercial stereolithogmoment, the main problem with prac- it may be appropriate to look more

it may be appropriate to look more raphy system (largest size, 1 meter), tical autostereoscopic systems is that carefully again. Good examples are

carefully again. Good examples are although frankly this only had a very the viewing position can be critical. recent research prototypes by NHK, weak connection with other aspects of One American remarked to me that he one of a 50-inch autostereoscopic 3D this symposium. One of the most impresdidn't understand why there was so display using a 1.5 Mx3 pixel projector sive medical applications was from much interest as there seemed to be (1440x1024 resolution) and an air-filled

(1440x1024 resolution) and an air-filled researchers at Chiba University Hosmajor technical problems and, in fact, lenticular screen and the other a 9-inch pital, whose goal is to provide a tool there might even be a wall that could 3D electroluminescent display. Also,

3D electroluminescent display. Also, “with which many doctors can, with not be breached. Another commented Hamasaki's photographic work clearly their own eyes, simultaneously and that he found the difficulties encoun- indicates that the viewer positioning clearly observe a 3D diagnostic image tered when moving from viewing lobe problem can be solved, or very signifi- of an object organ from any direction to viewing lobe (i.e., head movement) cantly reduced. Even in the case of which is selectable and controllable by in glassless lenticular systems to be far restricted viewer position, there are the doctor's hand." For some years the more problematic than properly pre- obvious applications, such as sitting in stereoviewer has been available for sented glasses approaches. A third said front of a computer monitor looking at use with angiographic images, but these that he saw no Japanese systems that the image of a molecule. Also, the inter- cannot be viewed by several doctors were anywhere near being productiza- est in video conferencing is very great simultaneously. Similarly holographic ble and some that were much more in Japan. As one participant put it, you images are sometimes used, with all than a decade away.

can have a successful company selling their current difficulties. Of course, 2D Two Westerners expressed serious shutter glasses, or you can make real displays of synthesized computed tomogconcern that many of the scientists they money designing video conferencing raphy (CT) and magnetic resonance met from Japanese companies (espe- for businesses. Finally, whoever said imaging (MRI) slices are also widely cially the younger ones) had huge gaps that the Japanese were daunted by used. The Chiba experiment uses x-ray in their knowledge of what had already 10-year time frames?

images recorded on a video tape or been done, and that a number of things At least one speaker H. Tilton 8-mm camera. Images are generated were being worked on from scratch (Visonics) disagreed with the claim that via a total circumference radiograph again, albeit with new materials. For 3D HDTV was far away. He felt that a system (circa 1982) that can take 240 example, one attendee commented to very modest increase in HDTV chan- images in 8 seconds over 360°, i.e., with the authors that the NTT eye-tracker nel space compared to 2D transmis- a 1.5° pitch. The current system takes (above) had been done at Hewlett- sion was all that was needed, and that if

sion was all that was needed, and that if these images, enhances them, and makes Packard 10 years ago.

an appropriate standard was adopted a them available 12 at a time in a frame buffer, which then displays them via a When I wrote about virtual reality TOURS CRT with lenticular screen with 1 mm last year (see reference above), I menpitch and 12 dots per pitch (correspond- tioned work by S. Tachi (University of As part of the Symposium on 3D ing to 18 degrees of rotation). The Tokyo). Prof. Tachi showed a very Imaging Technology and Arts, the authors are hoping to improve the system impressive movie at the symposium organizers arranged for foreign particin the future, but they note that a major based on research done over more than ipants to visit four laboratories in the difficulty is that there are no currently 10 years in Japan. His work has focused Tokyo area to see representative work available flat lenticular glass sheets on developing anthropomorphic robots in this field. Visits were short and installed within the glass envelope of a that will work in synchrony with a focused. The description below is meant CRT and that they are looking into human (so-called master-slave system). only to give a rough sense of what we using LCDs.

The operator, using a helmet and data saw and put it in perspective. One of the most obvious applica- glove, moves around and the remote tions of 3D is to remote sensing, photo- robot duplicates his movements. (This NHK Science and Technical grammetry, and geographic informa- concept is not new, but Tachi's work Research Laboratories tion systems. S. Murai gave an exciting seems to be one of the best implemenpaper of what can be done, automati- tations.) The weight of the visual dis- This laboratory celebrated its 60th cally, from digital elevation models play is carried by a link mechanism with anniversary in 1990. Three hundred (DEMs). DEMs can be obtained from six degrees of freedom, which cancels twenty people (270 research staff) are stereo aerial photography, stereo space gravitational force by use of a counter working in three general areas related imagery, and rasterized contour maps. balance mechanism. The master arm to broadcast engineering. The annual Each of these requires different tech- has 10 degrees of freedom; 7 degrees of budget is about ¥7.7B ($62M), approxnology and Murai listed various proj- freedom are allocated for the arm and imately 1.3% of NHK's total budget. ects that groups in Japan have been an additional three are used to comply General research thrusts are as folworking on. For example, image match- with the body movement. The operator ing (which is needed for aerial and has input to his goggles not only from space images) is being studied by the TV, etc., mounted on the robot, but • Putting new broadcast media into Japan Society of Photogrammetry and also from integrated computerized practice. In August 1990 the third Remote Sensing, a research commit- displays that allow him to “see” what Japanese broadcast satellite was tee set up by MITI, the Japan Highway the robot is seeing even if the robot's launched, BS-3. HDTV is being Corporation, the Geographic Survey vision is obscured, for example, by smoke. transmitted from this 8 hours each Institute, etc. The key to progress in Thus, this system assumes a known envi- day as well as some data-broadcasting this area is how to automate the digital ronment to work in. We watched a NHK has been working on HDTV data. While not discussing any details demo of the robot lifting eggs from a since 1964 and demonstrated a stereo of this work, Murai gave several fas- basket as well as turning off a valve in a HDTV system in 1989. We were cinating practical examples including smoking room. The most interesting told that there is no plan yet for a DEMs generated automatically by image aspect of this demonstration was that commercial product, but one is matching of stereo aerial photos and Tachi assured the audience that the expected soon. Broadcasting equipthen used for a highway plan, 3D drain- operator needed no training to use the ment, transmission systems, and age maps generated automatically, a system because it operated very natu- home receivers are studied at the bird's eye view of a shaded hillside in rally. He put it more dramatically: laboratory. Currently it is possible Bhutan obtained by processing an input operators already come to the system to purchase for home use a 50-inch contour map, a 25-second 3D animation pretrained by living.

HDTV monitor for about $25K (in "flying" along a fault line in Shikoku The "art" portion of the sympo- Japan this is called High Vision TV) Island and, finally, a “fly-by" of Mountsium was provided by J. Gjessing that will allow viewing of prerecorded Fuji. I think that everyone in the audi- (Norway), who demonstrated various tapes from a VCR or viewing of ence was very impressed with the fact aspects of his anaglyphic slide work. decoded signals from the satellite. that this could be done essentially Conference attendees also had the However, the satellite signals need automatically, as well as with the infor- opportunity to learn about “virtual to be decoded. At the moment this mation content of the images.

drugs” at an evening session. I don't requires another box (MUSE think that any of the Westerners decoder) about the size of a VCR, attended.

about $10K. We were told that an probably have the same orientation in tion and will be effective as a sales tool.

implementation via large scale inte difficult part is how to encode all the The final tour stop was a standards gration (LSI) was in progress. The different phase information on a tape laboratory, where a set of "standard” laboratory is also working on a or disk and then distribute it to the HDTV still pictures was being investi33-inch, wall-mounted, plasma dis- speakers properly. Wedid not have any gated for various characteristics such charge panel (6-mm-thick) HDTV opportunity to discuss this aspect. We as color, sharpness, etc. Standard moving display and is heavily involved in were given a demonstration of field images have not yet been selected. Our studio and international standards. sequential stereo HDTV based on film host at NHK, who also presented a

(not computer generated). In principle paper at the symposium summarizing • Improved conventional broadcast this is fairly direct, but the implemen- the company's work, was Dr.

services. Better reception, noise and tation details are difficult and NHK Hideo Kasaka. interference reduction research, such pulled it off very well. Everyone was as optical cable TV. Work on impressed with the visual effect; bright- Matsushita Electric Works machine translation (as well as some ness, clarity, and depth were excellent. speech and pattern recognition), In fact R. Kroiter, who is one of the This virtual reality exhibit was one which is already in use for making founders of Imax, commented that the small part of a large public exhibit area subtitles for news programs. New apparent resolution was even greater

apparent resolution was even greater where many of the company's consumer cameras for use under a very wide than what he normally expected even products were on display including range of light (sunlight to starlight) from HDTV. (See also the remark by such low-tech ones as light bulbs. The and wireless (45-GHz) broadcast Starks above about information con- VR system incorporated a conventional cameras. Experimentation with tent in stereo images.) We were shown head-tracker, goggles, data glove, and synthesizing HDTV and real-time a demonstration of the autostereoscopic stereo headphones. As usual with images using “chromakey” (see also system based on a 50-inch lenticular these systems, the LCD resolution in discussion of Sony below). There is projection system. The resolution is the goggles was not as good as one also a very interesting experiment not as good as the HDTV system, and would like, although it was about as with digital broadcasting that allows viewers had to be carefully seated, but good as one normally sees. Also, several use of 35- to 40-cm-diameter anten- no glasses were needed and the image ofthe participants complained that the nas for nearly all of Japan reception was amply bright and clear.

stereo video effects could be done better and even mobile reception.

One of the most interesting demon- and that the attendant who moved

strations (and the furthest into the around their hand was too intrusive. • Research into future broadcast sys- future) was by Mahito Fujü in the Visual Nevertheless, the most interesting part

tems. Materials for signal process. Science Research Department. Fujii's of this demonstration was that it was ing, recording, display, and LSI are work is directed toward developing a included as part of a kitchen planning studied. Also ISDB, Integrated fully 3D image beginning with only two system. The user is placed inside a kitchen Service Digital Broadcasting, the or three static 2D TV pictures. The that he/she can walk through and, by satellite equivalent of ISDN, which resulting 3D image can be rotated (a moving the data glove and “grasping,” will allow broadcasting of TV, audio, little). In other words, occluded infor- can open cabinets, turn on faucets, radio, fax, software (data), telemusic, etc., mation is interpolated as needed. His etc. Carping from the experts aside, I as well as 3D video and audio are approach is to use the fact that edges thought that it was a good demonstramajor projects.

the two images. He has built a four- Matsushita claims that it is the first We saw several examples on layer neural network to extract binocu- commercial application of VR. enhanced reality work at NHK, One lar disparity by the use of edge inforwas a wall of 160 small speakers that mation and the model can specify relative Sony Media World could be driven at appropriate phases positions in 3D space that can then be so that the sound appeared to emanate interpolated to form textured surfaces. We spent one afternoon at Sony's from a point several meters in front of Fujii admitted that this was still in the public display area in which a large the wall. NHK scientists told us that very early stages; his neural model runs number of its fanciest products and they could also localize (in space) dif- on NHK's Convex, much slower than prototypes were on exhibit. This included ferent kinds of sounds, such as differ- real time. (I was told that there is also a multipurpose teleconferencing room, ent instruments. This type of research related work at the Heinrich-Hertz HDTV displays (both 2D and field has been done in other places. The Institute in Germany.)

sequential), and an extremely sophisticated intelligent TV studio where just about every digital processing technique was available from within a "videocockpit" to control production, editing, and transmission. This included chromakey, converting 2D images to 3D, and others. In the stereo TV demonstration we were shown two films composed of still photos of Japanese museum pieces and a third (much more effective) showing a training film for ophthalmic surgery. The latter revealed just how helpful the sensation of depth can be in transmitting information. In both this demo and in NHK's, it was pointed out just how difficult the photographing of multiple images can be and how much care must be exercised during all steps of the process. Sony's exhibits, like Matsushita's, covered much more than we were able to see during one afternoon.

« iepriekšējāTurpināt »