rich@net1.ucsd.edu (bmf) (07/29/89)
I have written software for color desktop scanners for about three
years. In that time I have yet to see one that I really feel
positive about. Most of them have some good points but...
Anyway this discussion is for hobby purposes. The software I write
for a living but the hardware a do at home for fun like any Buckaroo B.
fan.
Problem:
Creating a desktop color scanner which is easy to use, has
the quality of a drum scanner , and costs less than $3000 in
parts. If we keep the discussion public we all gain a public
domain scanner.
Todays Scanners
Todays scanners rely on either CCD arrays or video chips to
collect a scan line, or block of pixel data. Both of these
devices suffer from dynamic range problems. Determining black
and white on them isn't that bad but getting a linear
response is. Maybe a different technology should be applied
here.
Both of these devices need to be moved around the image (or the
image moved around device) This means adapting some sort of
stepper motor technology. This leads a loss of sharpness in
the image. Some companies handle it better than others, but
again maybe a different sort of solution is avaiable.
Software solutions tend to be slow and make an image more
crisp rather than match the original.
The CCD technology relies on either color light bulbs or
color filters to reproduce the color in the image. Nasty
problems occur here. The spectral response of both the
filters and the bulbs tend to look like overlapping upside
down u's. So if you are using red, green, and blue filters,
there are certain reds greens and blues which do not
register at all. While there are some greens that come
register as yellow and others that register as cyan. So a
green grass may effectively dither out to a grey. (this
really pisses off scanner salespeople if you scan an image
like this in front of them at a trade show.)
Both the CCDs and the Video chips rely on camera lenses to
focus the image unfortunately a red, green, and blue light
have different focal lengths, so again we have a small
sharpness problem.
Conclusion:
You can do a good job with todays scanners but what about
applying some other technologies. Are there CD lasers
available in a variety of wavelengths? How about focusing
one of the spectral analysis devices onto a ccd array and
any diode which shows a response gets added into the color
value of the current pixel being scanned (too slow huh?).
This could be fun.
-Rich
/* Rich Stewart {dcdwest,ucbvax}!ucsd!net1!rich */root@cca.ucsf.edu (Systems Staff) (07/29/89)
The Chinon scanner that I saw at the deNardi (Norm, I apologize
if I didn't get the capitals right) show a couple of months ago
looked interesting.
It's cheap ( < $1000 ) scans a fixed page so filtered light scanning
wouldn't have the registration problem that a moving paper device
would. The image being scanned is out in the open to make filtered
lighting easy.
And the documentation tells how the interface (rs232) protocol
works so you can write a driver.
Has anyone used this device? I would like to see some reports
of experience with it.
Thos Sumner Internet: thos@cca.ucsf.edu
(The I.G.) UUCP: ...ucbvax!ucsfcgl!cca.ucsf!thos
BITNET: thos@ucsfcca
U.S. Mail: Thos Sumner, Computer Center, Rm U-76, UCSF
San Francisco, CA 94143-0704 USA
OS|2 -- an Operating System for puppets.
#include <disclaimer.std>cliff@ficc.uu.net (cliff click) (07/31/89)
In article <1869@ucsd.EDU>, rich@net1.ucsd.edu (bmf) writes: > Problem: > Creating a desktop color scanner which is easy to use, has > the quality of a drum scanner , and costs less than $3000 in > parts. If we keep the discussion public we all gain a public > domain scanner. Maybe you should do the reverse of a laser-printer. Scan the image with a laser reflected from a rotating mirror (to sweep the page), and check the intensity of the reflected light. Use different colored lasers (or filters on a white-light laser??) for different colors. You probably need to correct the recieved data for the spectral response of the laser/reciever. Maybe you could use a standard laser-printer/xerox shell (motor, paper transport, laser works) and add your own electronics. -- Cliff Click, Software Contractor at Large Business: uunet.uu.net!ficc!cliff, cliff@ficc.uu.net, +1 713 274 5368 (w). Disclaimer: lost in the vortices of nilspace... +1 713 568 3460 (h).
childers@avsd.UUCP (Richard Childers) (08/03/89)
rich@net1.ucsd.edu (bmf) writes: > Problem: > Creating a desktop color scanner which is easy to use, has > the quality of a drum scanner , and costs less than $3000 in > parts. If we keep the discussion public we all gain a public > domain scanner. . . . > Conclusion: > You can do a good job with todays scanners but what about > applying some other technologies. Are there CD lasers > available in a variety of wavelengths? How about focusing > one of the spectral analysis devices onto a ccd array and > any diode which shows a response gets added into the color > value of the current pixel being scanned (too slow huh?). > This could be fun. This is all very interesting, as I was peripherally involved in such a project several years ago. An individual whom was peripherally involved in printing, and some shadowy partners, wanted to be the first people to provide tabletop printing-quality-resolution scanning and reproduction systems. I provided the organizing element to an otherwise incoherent plan to become millionaires, such that the algorithm we evolved after much discussion resembled yours in approach, although we foundered long before we reached the point of actually researching the interesting details, for lack of responsible funding. Getting a scanning resolution of 1/1000th of an inch was the major problem. For some reason, the senior financing partner wanted to have everything written in PostScript, but properly speaking, the accounting and control software would have been written in another language, more suited, and the PostScript would have been restricted to the image-processing module. There were three problems. (1) Create a device which attaches to a CPU, via SCSI, or RS232, or whatever, that translates a physical image into a series of bitmaps corresponding to the various major colors. (2) Create and implement algorithms suitable to translating the major colors, from the physicist's point of view, into the complementary color set, as seen from the ink- driven view of the printer. (3) Create and implement algorithms to translate vast bitmaps into a series of concise and minimalist PostScript language descriptions of the image contained, perfectly. This particular thread represents a discussion on problem (1). ( When we were talking about this, I was pretty sure monochromatic lasers were the light source to use, and I'm pleased to see my guess confirmed. ) I've seen many discussions that address the issues around problem (2), and I've learned quite a lot that I was unable to ferret out of libraries, by the way, while reading them. Problem (3) doesn't seem that hard to me, sort of a compiler problem, but instead of dealing with textual icons you'd be dealing with geometric ones, probably hexagons and triangles to fill up all the available spaces, then lines where possible, then point-specific statements, the idea being to crunch the huge bitmap down to the smallest PostScript description possible, unless time constraints forced a tradeoff between time and space. ( A small PostScript description would pay off in the reproduction phase, though, or so I assume, perhaps naively. ) There's also a level somewhere in here, probably 1.5, where after you've scanned in the image, but before you've separated it, where you might find it convenient to scale it, crop it, rotate it and in general engage in an additional layer of layout operations before submitting the job for printing. This would require a high-density monitor, though, and even the best of the monitors available for the Mac II ( we got one as a potential platform ) was inadequate for the job, realistically speaking. This may still be a hitch to serious layout and graphic design software, by the way. It's good resolution, but not _great_ resolution, not _invisible_ resolution. I've been waiting for two or three years for something to appear on the market, but it seems to be the same old 300 DPI laser writers. < Sigh. > I don't see a public domain scanner getting very far - hardware efforts need physical proximity or some serious self-discipline on the part of multiple individuals - but I'm interested in seeing this happen, and I'm interested in helping, if it comes to some genuine development efforts. >/* Rich Stewart {dcdwest,ucbvax}!ucsd!net1!rich */ -- richard -- * Truth : the most deadly weapon ever discovered by humanity. Capable of * * destroying entire perceptual sets, cultures, and realities. Outlawed by * * all governments everywhere. Possession is normally punishable by death. * * ..{amdahl|decwrl|octopus|pyramid|ucbvax}!avsd.UUCP!childers@tycho *
hutch@celerity.uucp (Jim Hutchison) (08/05/89)
In <1869@ucsd.EDU> rich@net1.ucsd.edu (bmf) writes: [...] > Anyway this discussion is for hobby purposes. The software I write > for a living but the hardware a do at home for fun like any Buckaroo B. > fan. Sets the footing as non-print quality scans, perhaps. What kind of resolution are we talking about here? Drums range up to 3-4 digits to the inch and are generally too large to put on desktops. [...] > The CCD technology relies on either color light bulbs or > color filters to reproduce the color in the image. Nasty > problems occur here. The spectral response of both the > filters and the bulbs tend to look like overlapping upside > down u's. So if you are using red, green, and blue filters, > there are certain reds greens and blues which do not > register at all. How about using a different color scheme, such as YIQ. N passes, based on the response curves of your lights (3 in the case of YIQ). Likely this would give you shaky color passes and a good grey pass. Now just because YIQ is what's in NTSC doesn't mean that you have to use them at low bandwidth like NTSC, the bandwidth of the I(cyan->red) and Q(yellow->blue). YIQ is designed around human vision, which also has response curves like "overlapping upside down u's". For a pleasant bit of data on YIQ, see "Illumination and Color in Computer Generated Imagery" by Roy Hall. [...] > Both the CCDs and the Video chips rely on camera lenses to > focus the image unfortunately a red, green, and blue light > have different focal lengths, so again we have a small > sharpness problem. I guess this could also be seen as a resolution limit? Isn't this more of an aperature problem? I can see wanting a larger aperature for a faster scan. If you need the resolution/focus, couldn't you just crank down the aperature for one or two of the passes? /* Jim Hutchison {dcdwest,ucbvax}!ucsd!celerity!hutch */ /* Disclaimer: I am not an official spokesman for FPS computing */
poynton@vector.Sun.COM (Charles A. Poynton) (08/08/89)
Jim Hutchison <hutch@celerity.uucp> says: > How about using a different color scheme, such as YIQ. N passes, based > on the response curves of your lights (3 in the case of YIQ). I and Q, and the related pairs (U and V) and (B-Y and R-Y), are signals that are ubiquitous in colour television, because the bandwidth of these signals can be dramatically reduced compared to that of luminance without visual effect. These signals are all colour-difference signals, which means that they are formed by the subtraction of luminance from something (for example, blue minus luminance). Each is bipolar, that is, goes negative for some substantial portion of the spectrum. No optical filter can be used to extract such a signal from a scene, because all optical filters are based on absorbing, reflecting, or transmitting photons. For practical purposes, there are only positive contributions from photons, and negative lobes are out. You can synthesize negative lobes electronically, but you can't build an optical filter to do it. Unless you know really a lot about scanners, stick to RGB of some sort. C. ----- Charles A. Poynton Sun Microsystems Inc. <poynton@sun.com> 2550 Garcia Avenue, MS 8-04 415-336-7846 Mountain View, CA 94043 "Japan has no laws against damage to its flag, but it has strict laws forbidding the burning of foreign flags lest this give offense to the country in question." -- The Economist, July 1, 1989, p. 19. -----
paj@hrc63.uucp (Mr P Johnson "Baddow") (08/08/89)
In article <5368@ficc.uu.net>, cliff@ficc.uu.net (cliff click) writes: ... > the intensity of the reflected light. Use different colored lasers (or > filters on a white-light laser??) for different colors. You probably ... There is no such thing as a white laser. Lasers produce monochromatic light. White light is a mixture of all frequencies. You would need three lasers and a mechanism for selecting them. The rest of the idea, a laser printer based scanner, is pretty neat. Would a parallel beam of white light work OK? Could it be accurate enough? What about refraction errors and fringing in the optics? What about filtered white light? How big is the average laser printer laser? What about mounting a photo-receptor in place of the laser and generally illuminating the document? -- Paul Johnson, | `The moving finger writes, And having writ, moves on,' GEC-Marconi Research | Omar Kyham when contemplating `vi'. ------------------------------------------------------------------------------ The company has put a radio inside my head: it controls everything I say!
saj%yipeia@Sun.COM (Scott A. Jordahl) (08/10/89)
In article <658@hrc63.uucp> paj@hrc63.uucp (Mr P Johnson "Baddow") writes: >In article <5368@ficc.uu.net>, cliff@ficc.uu.net (cliff click) writes: >... >> the intensity of the reflected light. Use different colored lasers (or >> filters on a white-light laser??) for different colors. You probably >... > >There is no such thing as a white laser. Lasers produce monochromatic light. WRONG!! White light lasers DO exists. I've used many in my time. Unfortunately, most white light lasers are krypton-ion and are about a meter in length, are water cooled, and consume LOTS of power (~60 amps). Except for these minor details, this type of laser would make a GREAT basis for a desk top scanner. :-) The white light is indeed a mixture of all colors of the spectrum and can be broken apart with the use of a prism. >-- >Paul Johnson, | `The moving finger writes, And having writ, moves on,' >GEC-Marconi Research | Omar Kyham when contemplating `vi'. >------------------------------------------------------------------------------ >The company has put a radio inside my head: it controls everything I say! /########################################################\ | Scott A. Jordahl | | UUCP: saj@yipeia.sun.com | | PHONE: WK: [415] 336-5463 | | HM: [408] 270-5619 | \########################################################/
brian@hpfcdj.HP.COM (Brian Rauchfuss) (08/10/89)
Using three monochromatic lasers has a problem: only
colors containing the three wavelengths wavelengths of
the lasers would show up. A color which only reflects
a narrow bandwidth in-between the colors of the lasers would
show up as black! (For example, a monochromatic yellow would
be totally missed by monochomatic red light and monochomatic
green light).
What you need for the scanner is relatively broad-band light
(using filters on white light maybe) so that you can determine
intermediate colors by measuring low level reflections of two
input lights.
-----------------------------------------------------------------
Brian Smokefoot "I never knew I could shape my life
brian@hpfcbdr.HP.COM like the artist paints his dreams
on a canvas." - Minor Detailmcdonald@uxe.cso.uiuc.edu (08/11/89)
>>There is no such thing as a white laser. Lasers produce monochromatic light. >WRONG!! White light lasers DO exists. I've used many in my time. >Unfortunately, most white light lasers are krypton-ion and are about a >meter in length, are water cooled, and consume LOTS of power (~60 amps). >The white light is indeed a mixture of all colors of the spectrum and can >be broken apart with the use of a prism. "White" krypton lasers are not really white. They contain several lines that add together to appear whice to the eye. They do make good color scanner sources, as the lines are in nice places. The only really white lasers are the very, very shortest pulsed femtosecond ones, or the continuum generated by focusing short pulses into water. This is not really a comp.graphics topic, but it is interesting to people intereted in things like scanners. Doug McDonald
sparks@corpane.UUCP (John Sparks) (08/14/89)
In article <46900035@uxe.cso.uiuc.edu> mcdonald@uxe.cso.uiuc.edu writes: > >"White" krypton lasers are not really white. They contain several >lines that add together to appear whice to the eye. They do >make good color scanner sources, as the lines are in nice places. > I am not an expert on lasers, but I came across some info that might be of interest. On the Discovery Channel there was a piece on medical lasers. Some surgeons needed different frequencies of lasers for different tasks, so they were using a Dye laser. From what I could gather, a primary laser excited a dye which emmitted a secondary laser which could be tuned to almost any frequency. Using such a laser maybe you could tune it to Red, Green, or Blue for color scanning. -- John Sparks | {rutgers|uunet}!ukma!corpane!sparks | D.I.S.K. 24hrs 1200bps ||||||||||||||| sparks@corpane.UUCP | 502/968-5401 thru -5406 Help fight continental drift.
brian@hpfcdj.HP.COM (Brian Rauchfuss) (08/16/89)
sparks@corpane.UUCP (John Sparks) writes: >Some surgeons needed different frequencies of lasers for different tasks, so >they were using a Dye laser. From what I could gather, a primary laser excited >a dye which emmitted a secondary laser which could be tuned to almost any >frequency. A dye laser is pumped by a primary laser and then lases at its characteristic frequency. The laser can be fine tuned around this frequency, but to get a big change in frequency you need to change the dye! Dyes are availible for nearly any color. (Since a dye lasers are relatively cheap, could you have three of them? or three dye tubes that use the same mirrors and pump?) ---------------------------------------------------------------------- Brian Smokefoot "... never knowing I could shape my life brian@hpfcbdr.HP.COM like the artist paints his dreams on a canvas." - Minor Detail
rich@net1.ucsd.edu (bmf) (08/22/89)
First of all, thanks for the good response. I am working on a
summary of what I have read so far and what I saw at the SPIE show.
But about this laser stuff....
I have been told a white light laser does not exists because
cramming all those frequencies together they start cancelling
out and moving in different directions after some distance.
Tunable lasers are neat, but if they skip any decent size part of
the visible spectrum you loose that color in the image and if it
takes more than a micro second to change frequency, you'll be
scanning for an awfull long time.
Something along the lines of defraction gradings aimed onto a ccd
to read frequencies or other spectral analysis tools may prove
interesting.
-Rich
/* Rich Stewart {dcdwest,ucbvax}!ucsd!net1!rich */mcdonald@uxe.cso.uiuc.edu (08/25/89)
>I have been told a white light laser does not exists because >cramming all those frequencies together they start cancelling >out and moving in different directions after some distance. White light lasers, very genuine ones, really do exist. There is one running right now about 30 feet above my head. The different colors do come out at different TIMES - but, do you care that one color comes out 0.0000000000001 second before another? In any case, further work is expect to get them all coming at out once anyway. Another factor of three faster pulses and there would indeed be genuine, all at once, white light. All the colors would come out in one pulse 0.0000000000000025 second long. I should add that these lasers are really something to behold. They generate lots of oohs and aahs from the non-jaded person - laser tables 5x14x2 feet weighing 5 tons crammed full of hundreds of optical parts. (This include amplifiers.) Doug McDonald
sleat@sim.ardent.com (Michael Sleator) (08/26/89)
In article <46900036@uxe.cso.uiuc.edu> mcdonald@uxe.cso.uiuc.edu writes: > > >>I have been told a white light laser does not exists because >>cramming all those frequencies together they start cancelling >>out and moving in different directions after some distance. > The participants in this discussion would do themselves, eachother, and all of the observers a favor by defining their terms. "White light" is a very ambiguous term in this context. We seem to have at least two conflicting interpretations in progress. One seems to be "radiation with a more or less continuous spectrum", and the other seems to be "radiation with sufficient spectral components that it appears white", or something like that. The indirect quote above, referring to "all those frequencies", suggests a continuous spectrum. The following paragraph seems to be referring to a "multi-line source"; one that produces multiple simultaneous (or nearly so) narrow-band outputs that viewed together appear white. From my limited knowledge of lasers, I can well believe that there are no continuous-spectrum lasers. How would such a beast work? In what sense would it even be a laser? (In my understanding of them, lasers inherently involve resonance, a notion rather at odds with a continuous spectrum.) (These are rhetorical questions. If you want to answer them, I'm quite sure this is *not* the place to do it. If you do it somewhere else, send me a pointer, please.) >White light lasers, very genuine ones, really do exist. There is one >running right now about 30 feet above my head. The different colors do >come out at different TIMES - but, do you care that one color >comes out 0.0000000000001 second before another? In any case, further >work is expect to get them all coming at out once anyway. Another >factor of three faster pulses and there would indeed be genuine, >all at once, white light. All the colors would come out in >one pulse 0.0000000000000025 second long. > >I should add that these lasers are really something to behold. They >generate lots of oohs and aahs from the non-jaded person - laser >tables 5x14x2 feet weighing 5 tons crammed full of hundreds of optical >parts. (This include amplifiers.) I'm sure they are, but it doesn't sound as if they're particularly appropriate for the original application: an optical scanner. To corroberate your point that multi-line lasers do exist and to bring it a little closer to home, I offer the following: In LASER FOCUS/ELECTRO-OPTICS, vol. 24, no. 12, Dec. 1988, p. 62 there is an advertisement by Nihon Dempa Kogyo Co., Ltd. ("NDK") for their "RGB LASER(tm)". This is a linearly polarized Helium-Cadmium laser, model RGB300, with the following specs: output wavelength RED GREEN BLUE (in nanometers) 636.0 537.8 441.6 635.5 533.7 output power > 25mW (total) beam diameter approx 1.0mm (1/e^2) beam divergence approx 0.6mrad (full angle) optical noise <1% The tag says, "NDK introduces its hollow cathode He-Cd laser commercially for the first time in the world. RGB LASER(tm) emits three primary colors, red, green and blue light, simultaneously and coaxially by a single laser tube." Model RGB400 is also available with 100-150mW output. They don't give any indication of the relative power outputs at the different wavelengths for either model. Note that these are CW lasers, not pulse lasers like the monster described above. It doesn't seem that the fact that the red and green outputs are twin lines would be a significant problem in a scanner application, since the pairs are quite close together. In the acompanying photo, they show the end of the enclosure, which appears to be a table-top box roughly the size one might expect for a comparable power He-Ne laser. The beam is shown being split into red, green, and blue components with a series of dichroic mirrors. It looks pretty good in the picture... No mention is made of cost. Nevertheless, it appears that the technology is there, and if a market exists, it can probably be made cost effective. (A thousand bonus points to the first person to report back with news of a scanner product actually using one!) Michael Sleator Ardent Computer 880 W. Maude Ave. Sunnyvale, CA 94086 408-732-0400 {apple, decwrl, hplabs, ubvax, uunet}!ardent!sleat