martin@beowulf.ucsd.edu (Bruce Martin) (04/05/88)
Has anybody had any experience with Perfect Vision? Is it vaporware? How does it compare in picture quality to DigiView? Is there any loss of quality using a color camera instead of a black and white one? Is it compatible with an amiga 1000? Any reviews and/or recommendations would be helpful. Anything you might have to say about genlocks would be appreciated as well. thanks, Bruce
doug@eris (Doug Merritt) (04/06/88)
In article <4834@sdcsvax.UCSD.EDU> martin@beowulf.ucsd.edu (Bruce Martin) writes: >Has anybody had any experience with Perfect Vision? Is it vaporware? >How does it compare in picture quality to DigiView? Is there any loss >of quality using a color camera instead of a black and white one? Is >it compatible with an amiga 1000? It's not vaporware...I've seen it on the shelf at HT Electronics. There will necessarily be a loss in *potential* maximum quality when using an NTSC color camera instead of a black and white w/ color wheel, because NTSC has very limited color bandwidth. I wouldn't know how close Perfect Vision comes to the optimum with the color wheel approach. They have two versions, one for the 1000, and one for 500 and 2000. They claim to grab a one bit plane image in 1/60 of a second, and imply N/60 seconds for color, where N is the number of bitplanes. I haven't used it nor seen it in operation; I'm curious as to the quality...it sort of sounds like the best of DigiView and LIVE! wrapped up in one neat package. BTW although a lot of work went into DigiView 2.0, and it produces much better quality images than 1.0 (or than LIVE!, of course), I am still quite disappointed in the lack of flexibility in fine tuning the results. They very obviously did not have an image processing expert work with them on this. To get really high quality results depends far more on how well you pick the subject matter, and how carefully you adjust lighting conditions, than it does on any software adjustments. Too bad...you can do a lot with 21 bits/pixel of image information. If you bother. Which they didn't. Other than relatively straightforward dithering, which is necessary but not sufficient. All of which is to say that, if the Perfect Vision folks wanted to outdo DigiView, it would be relatively easy. Doug Merritt doug@mica.berkeley.edu (ucbvax!mica!doug) or ucbvax!unisoft!certes!doug
ali@polya.STANFORD.EDU (Ali T. Ozer) (04/06/88)
In article <8379@agate.BERKELEY.EDU> doug@eris.UUCP (Doug Merritt) writes: >BTW although a lot of work went into DigiView 2.0, and it produces much >better quality images than 1.0 (or than LIVE!, of course), I am still >quite disappointed in the lack of flexibility in fine tuning the results. Last night at FAUG, Tim Jennison of NewTek announced DigiView 3.0. It'll apparently cost $15 to upgrade. He actually showed it too, and said something about 2 weeks... I didn't catch if that was "it'll go alpha in two weeks" or "it'll ship in two weeks." He also mentioned that the availability of almost everything else from NewTek would be summer --- This means DigiPaint 2.0 and the Toaster. DigiView 3.0 provides overscan, upto 352/704 x 240/480 resolution. He added that they also worked on the algorithms a bit to produce better colors, but didn't sound like it was a major change... Ali Ozer, ali@polya.stanford.edu
kim@amdahl.uts.amdahl.com (Kim DeVaughn) (04/06/88)
In article <8379@agate.BERKELEY.EDU>, doug@eris (Doug Merritt) writes: > > BTW although a lot of work went into DigiView 2.0, and it produces much > better quality images than 1.0 (or than LIVE!, of course), I am still > quite disappointed in the lack of flexibility in fine tuning the results. > They very obviously did not have an image processing expert work with > them on this. To get really high quality results depends far more on > how well you pick the subject matter, and how carefully you adjust > lighting conditions, than it does on any software adjustments. > Too bad...you can do a lot with 21 bits/pixel of image information. > If you bother. Which they didn't. Other than relatively straightforward > dithering, which is necessary but not sufficient. > > All of which is to say that, if the Perfect Vision folks wanted to > outdo DigiView, it would be relatively easy. Last night at the FAUG meeting, Tim Jennison <I've probably misspelled his name again>, New-Tek's President, announced that Digi-View v3.0 would be released in "about two weeks". Registered owners should receive upgrade information via a newsletter that's coming out. He did demo the new version, and said they had "tweaked" their algorithms some. Also, the new version will support overscan, and has a couple of other new features/capabilities. Not alot of time was spent on the New-Tek stuff, as the meeting was running a bit late, and Max Toy was up next, so I really don't have any more information, except that ... The Video-Toaster should be available "in the summer", as should the long awaited Digi-Adaptor (add-on box to let you do color digitization from a VCR, etc). Also mentioned was Digi-Paint 2.0. Availability is ... you guessed it ... this summer. As a side note, if you're knowledgable about imaging-processing techniques, you can always save the digitized images in RAW form, and then process them with your own routines. The RAW format is 24-bits/pixel, however only 21-bits/pixel are meaningful (7-bits/color). /kim -- UUCP: kim@amdahl.amdahl.com or: {sun,decwrl,hplabs,pyramid,ihnp4,uunet,oliveb,cbosgd,ames}!amdahl!kim DDD: 408-746-8462 USPS: Amdahl Corp. M/S 249, 1250 E. Arques Av, Sunnyvale, CA 94086 CIS: 76535,25
david@ms.uky.edu (David Herron -- One of the vertebrae) (04/07/88)
Digi-View sounds like a neat product, and it certainly produces some very nice pictures. I like the fact that the resolution is 21 bits per pixel -- that's a neat feature. As I understand it however it takes a fair bit of time (>1 second?) to do each pass at the color wheel and that there are 3 (or 4) passes on the color wheel to make a picture. I would really prefer to be able to digitize frames on the fly, which leads one to Live. But Live doesn't work on my 2000 (I assume there'll be a version eventually which'll live on the bus). Also I don't know what sort of resolution Live has. I know that it doesn't do a very good job of freezing frames -- I've seen examples at the local store where they were digitizing frames from Predator and the moving bushes came out fairly blurry. Perfect Vision does no better -- color digitizing has to be done in >1 passes and doesn't have enough resolution. The Toaster will probably do exactly what I want but I can't afford $800-$1000 for it. Is DigiView my only choice? -- <---- David Herron -- The E-Mail guy <david@ms.uky.edu> <---- or: {rutgers,uunet,cbosgd}!ukma!david, david@UKMA.BITNET <---- <---- I don't have a Blue bone in my body!
ggibeau@ucqais.uc.edu (George Gibeau) (04/09/88)
In article <8379@agate.BERKELEY.EDU>, doug@eris (Doug Merritt) writes: > In article <4834@sdcsvax.UCSD.EDU> martin@beowulf.ucsd.edu (Bruce Martin) writes: > >Has anybody had any experience with Perfect Vision? Is it vaporware? > >How does it compare in picture quality to DigiView? Is there any loss > >of quality using a color camera instead of a black and white one? Is > >it compatible with an amiga 1000? > > It's not vaporware...I've seen it on the shelf at HT Electronics. > It does indeed exist. A friend of mine purchased it for their lab, but had some initial problems. It only seemed to work with one very old camera. They tried multiple cameras, including a brand new TTY (?) from Hitachi. After trying to no avail, the unit was send back to the factory to be looked at. It returned with a brief note: Try a shorter cable. Sure as heck, it worked perfectly when the cable was shortened from about 6 feet to about 1 foot. Problem number 2--- ******* it will NOT work with an 020 chip installed in a 1000. Upon yanking the 020 chip out of the 1000, it worked fine. Also, the same unit worked perfectly on a 2000 (again with no 020 chip). It does loose a little in crispness of image, but it is a heck of a lot faster and not subject to movement (as DigiView is). George Gibeau -- (Define the Universe and give 3 examples.) Idiots? They're worse than idiots. They're bureaucrats!! UUCP: uucba!ucqais!ggibeau BBS: (513) 721-7977 GT NODE: 006/005 US Snail-Dept of Biology ML 06, University of Cincinnati, Ohio 45221
doug@eris (Doug Merritt) (04/10/88)
In article <1412@ucqais.uc.edu> ggibeau@ucqais.uc.edu (George Gibeau) writes: > It does indeed exist. A friend of mine purchased it for their >lab, but had some initial problems. Thanks for the info. Can you get your friend to give us the technical specs? For instance, how many bits per pixel can it grab for maximum quality (does it compare to DigiView's 21 bits per pixel?) For non-ham modes, does it framegrab at exactly ((# bit planes) divided by 60) ? Does it support HAM at all? Do they, perchance, document how to directly read their interface themselves? This last point has been a burr in my side for a long time...it'd be nice to be able to write my *own* software to grab DigiView images. Similarly for sound samplers. Future Sound was kind enough to send me examples of C code to do so for their sampler, but it'd be nice if we had a standard way to talk to *any* sound sampler. Like a "snd:" device, for instance. I've been meaning to write one for Future sound for the last 15 months but it doesn't look like I'm going to get around to it after all. Doug Merritt doug@mica.berkeley.edu (ucbvax!mica!doug) or ucbvax!unisoft!certes!doug
kdd@well.UUCP (Keith David Doyle) (04/11/88)
In article <8532@agate.BERKELEY.EDU> doug@eris.UUCP (Doug Merritt) writes: >Thanks for the info. Can you get your friend to give us the technical >specs? For instance, how many bits per pixel can it grab for maximum >quality (does it compare to DigiView's 21 bits per pixel?) It grabs 4 bits per pixel, period. It has hardware contrast/brightness controls instead of software. >For non-ham >modes, does it framegrab at exactly ((# bit planes) divided by 60) ? It grabs a field at 4 bit planes in 1/60ths a second, a frame (two fields) in 1/30 sec. It will grab a still frame out of a moving video. >Does it support HAM at all? No, not with the current software, since it's all B&W. The promised NTSC to RGB converter would seemingly support HAM if the accompanying software does. >Do they, perchance, document how to directly >read their interface themselves? Yes, it comes with the C code to read frames. It's best feature is the ability to go on "automatic" and grab a series of frames (until you run out of RAM). You put your VCR in slow mode, and you can get a nice sequence. I'm hoping they'll add the ability to multi grab in a mode that does 1/4 screen frames, or even 1/9, 1/16th etc. so I can grab bunches of frames that were designed to be reduced anyway. Keith Doyle decvax!trwrb!cadovax!keithd
keithd@cadovax.UUCP (Keith Doyle) (04/13/88)
In article <8379@agate.BERKELEY.EDU> doug@eris.UUCP (Doug Merritt) writes: >There will necessarily be a loss in *potential* maximum quality >when using an NTSC color camera instead of a black and white w/ color >wheel, because NTSC has very limited color bandwidth. I wouldn't >know how close Perfect Vision comes to the optimum with the color >wheel approach. Hard to say. I agree that the BW of NTSC is a stumbling block and this holds true with videotape sources as well as from color cameras. And while I find PerfectVision useful, I would not give up my DigiView for still art. I find them each useful, according to what they're best at. DigiView for still art, PerfectVision for multi-frame capture from videotape. >They have two versions, one for the 1000, and one for 500 and 2000. Don't ask me what the difference is, the one I have has a standard centronics parallel connector on the back. I've used it on both a 1000 and 2000. >They claim to grab a one bit plane image in 1/60 of a second, and >imply N/60 seconds for color, where N is the number of bitplanes. I believe it's an entire grayscale (4 bits) image in 1/60th a second, with 3 shots to get RGB. Since it takes a couple of seconds to upload the image from the box, color grab must be done from a still image, while B&W is realtime. >I haven't used it nor seen it in operation; I'm curious as to the >quality...it sort of sounds like the best of DigiView and LIVE! >wrapped up in one neat package. I see that it addresses a need, to get decent frames from videotape. I don't see it replacing my DigiView, just fitting in in a slot where DigiView didn't perform. >BTW although a lot of work went into DigiView 2.0, and it produces much >better quality images than 1.0 (or than LIVE!, of course), I am still >quite disappointed in the lack of flexibility in fine tuning the results. >They very obviously did not have an image processing expert work with >them on this. To get really high quality results depends far more on >how well you pick the subject matter, and how carefully you adjust >lighting conditions, than it does on any software adjustments. I think that's a fact of life. Lighting is EXTREMELY important and something image processing only becomes a band-aid for. >Too bad...you can do a lot with 21 bits/pixel of image information. >If you bother. Which they didn't. Other than relatively straightforward >dithering, which is necessary but not sufficient. Sounds like you haven't figured out how to use it yet. I thought it was quite flexible in tuning the results. Two things that help to know are: 1. When you want to pump up the brightness/contrast, move the brightness and contrast sliders up TOGETHER. If you just turn up the contrast, you lose a little brightness. I've found it most effective to tune the two pretty much together. 2. If you're working with color, try turning the saturation WAY DOWN. I use it around -15, and don't lose much color, it just tones down the predominate color. If you have an image with too much red (or blue etc.), DON'T just turn down the red, the image will just turn green. Leave the red alone and back the saturation way off. >All of which is to say that, if the Perfect Vision folks wanted to >outdo DigiView, it would be relatively easy. They both have their place, I wouldn't want to do without either of them right now. Keith Doyle # {ucbvax,decvax}!trwrb!cadovax!keithd Contel Business Systems 213-323-8170
thad@cup.portal.com (04/13/88)
The REAL reason for DigiView 3.0 is to fix the serious bug in 2.0 reagrding self-modifying code. yes, DigiView's DigiDroid (the motorized color wheel) does NOT function on a 68020 Amiga with the cache enabled; one has to disable the 020's cache if you wanna use DigiDroid. Now, wasn't it in May 1985 the people from Amiga said "don't do self-modifying code"? I'd hate to tell you all the stuff that DOESN'T function on an 020 Amiga due to self-modifying code.
rgd059@Mipl3.JPL.Nasa.Gov (04/13/88)
In article <8532@agate.BERKELEY.EDU> doug@eris (Doug Merritt) writes: >Thanks for the info. Can you get your friend to give us the technical >specs? Disclaimer: I did some of the software for Perfect Vision (the color display section), so I hardly have an unbiased opinion. Sorry if you think this is commercialism, but you asked! >For instance, how many bits per pixel can it grab for maximum >quality (does it compare to DigiView's 21 bits per pixel?) It gets 4 bits per pixel, gray scale. For color, you get 12 bits. That's enough for most Amiga display purposes, since that's all the Amiga can display. You can adjust brightness and contrast via the hardware, so you don't need the extra bits for that. In order to digitize color, you must make three passes, like DigiView, but each of the passes is MUCH faster. It works with color or b&w cameras, VCR's, and TV's. Currently, color is done via optical filters, but there is a color splitter box coming RSN that will allow you to digitize color from a NTSC color source. With the box, it would take maybe 1.5 to 2 seconds to digitize a color picture (longer to display it, of course). Resolution is 320X200 or 320X400. Sorry, no hi-res. It would add immensely to the cost. If we get enough requests, though.... (but it'd be expensive). >For non-ham >modes, does it framegrab at exactly ((# bit planes) divided by 60) ? It grabs all 4 bits in one frame time (1/30 second interlace, 1/60 non- interlace). So, it will stop a moving image. Unfortunately, it takes about half a second to move the data into the computer (through the parallel port), so you can get about 2 frames/second burst mode. Color, of course, takes three passes, so you must have a still source. But, even with optical filters, it can take as little as 5 seconds to digitize color (how fast can you swap filters?). >Does it support HAM at all? Yep! (that's my baby!) It will convert to either fast HAM (no color registers), slow HAM (color regs used to greatly reduce fringing), and 32-color modes. You can also lock the palette so it won't change, and it calculates the best fit for each pixel. > Do they, perchance, document how to directly >read their interface themselves? There is some assembly source included that demonstrate how to read the hardware. I believe it's the actual routines used by the software. Good 'nuff? >This last point has been a burr in my side for a long time...it'd be >nice to be able to write my *own* software to grab DigiView images. DigiView is almost all software, so they'd have to give away most of their program (at least the meat of it). Perfect Vision is mostly hardware. >Similarly for sound samplers. Future Sound was kind enough to send >me examples of C code to do so for their sampler, but it'd be nice >if we had a standard way to talk to *any* sound sampler. Like a >"snd:" device, for instance. I've been meaning to write one for >Future sound for the last 15 months but it doesn't look like I'm going >to get around to it after all. Perfect Sound includes the complete source for the editor program (or at least an old version of it), including, of course, the digitizing source. If you get around to writing a device, we'd like to hear about it! Perfect Vision retails for $219, it's much less mail order. The color splitter should be about $80-90 range, but it hasn't been decided yet. (the splitter will also work with DigiView...) If you have any questions, you can call SunRize Industries at (409)846-1311. Or, you can email me... but I'm not an official representative of SunRize, I just do some contract programming for them on the side. (in other words, I can't sell you one ;-} ) Oh... someone in an earlier posting mentioned a problem with having to use a very short printer cable. That has been fixed, so the newer ones should work with reasonable-length cables. Bob Deen @ NASA-JPL Multimission Image Processing Lab rgd059@mipl3.jpl.nasa.gov span: mipl3::rgd059
doug@eris (Doug Merritt) (04/14/88)
In article <2038@cadovax.UUCP> keithd@cadovax.UUCP (Keith Doyle) writes: >And while I find PerfectVision useful, I would not give up my DigiView >for still art. I find them each useful, according to what they're >best at. DigiView for still art, PerfectVision for multi-frame capture >from videotape. Makes sense. I guess that leaves the question: how about PerfectVision versus LIVE! ??? > >I believe it's an entire grayscale (4 bits) image in 1/60th a second, >with 3 shots to get RGB. Since it takes a couple of seconds to upload >the image from the box, color grab must be done from a still image, while >B&W is realtime. Huh? If it takes a "couple of seconds" to upload an image, how could even B&W be realtime??? It doesn't actually have megs of memory in it, does it? [quoting me:] >>quite disappointed in the lack of flexibility in fine tuning the results. >> [...] To get really high quality results depends far more on >>how well you pick the subject matter, and how carefully you adjust >>lighting conditions, than it does on any software adjustments. > >I think that's a fact of life. Lighting is EXTREMELY important >and something image processing only becomes a band-aid for. Maybe I put that badly...yes, lighting is extremely important, and my results got 200% better when I switched from normal incandescent (yuk!) to photoflood bulbs. I'm still having problems sometimes with bright spots, though. Got a tip for that? They could fix *that* in software. First you shoot a grey background, which shows up the irregularities. Then DigiView could use that background map to adjust the image you shoot later to normalize the brightness. Easy, and useful. But you really *need* more processing than they give you...for instance, lasso an area and tell it, "use 8 extra colors just for this region", (useful for accurate rendition of, e.g. hair) lasso another and say "make this a uniform shade rather than trying to smooth shade it with dither" (useful for backgrounds). I guarantee you this feature would allow significant improvements to any picture you shoot. There are lots of other image-processing oriented features that would help, too. About the only spatial filter they include is "sharpness", which appears to be a high spatial frequency filter. >Sounds like you haven't figured out how to use it yet. I thought it >was quite flexible in tuning the results. I appreciate the suggestions about strategy, but I still want more features like the above. The problem still remains that all they really give you are global tuning controls that act like camera controls; it always helps to have some local selectivity for particular regions. I agree with you that what they provide is very useful, so far as it goes. Doug Merritt doug@mica.berkeley.edu (ucbvax!mica!doug) or ucbvax!unisoft!certes!doug
blubaugh@ucqais.uc.edu (Dwight Blubaugh) (04/14/88)
We bought Perfect Vision exactly for the reasons of incorporating an image acquire function in our software. The DigiView people were very uncooperative in telling us much about how their stuff works or even how the 24 bit stuff is stored. The Sunrize people have been more than cooperative in providing everything you need to roll your own. We got a bad one and it was replaced promptly. Also the short cable note is for real. Just don't use anything over 4 feet. I have lot's of people interested in the video separater (just in case anyone in Sunrize needs a hardware beta-tester :-) -- president of Ohio Valley Amiga Users Group (Cincinnati Ohio P.O. Box 428539) It's good to be king! UUCP: {decuac,mit-eddie,phri,pyramid}!uccba!ucqais!blubaugh USMAIL: Dept of Chemistry ML172, University of Cincinnati, Cincinnati OH 45221
fiddler%concertina@Sun.COM (Steve Hix) (04/15/88)
In article <8715@agate.BERKELEY.EDU>, doug@eris (Doug Merritt) writes: > > Maybe I put that badly...yes, lighting is extremely important, and my > results got 200% better when I switched from normal incandescent (yuk!) > to photoflood bulbs. > > I'm still having problems sometimes with bright spots, though. Got a tip > for that? Why not try (make) a light tent. You can get them commercially, but after looking at a few, it might be better to make your own. Basically, it's a tent made of white cloth (sacrifice a sheet or two) that encloses the subject and at least the objective lens of your camera. The lights stay outside the tent, shining on it, and the cloth diffuses the light very evenly across the subject. Without an effective point source of light, the hot spts go away. (Of course, it isn't likely to be perfect, since some of the light will go straight through the cloth rather than being completely scattered, but it should be a great improvement.) You'll probably have a drop in light intensity on the subject of about one f-stop, maybe two, depending on how thick the tent is. I built one to cover a fairly large object once out of PVC water pipe and PVC fittings (pressed together, no glue), and hung a couple of sheets over the frame. Worked fine. Also used all the lights I had at the time. And most of the living room. Kids loved it. seh
rusty@hocpa.UUCP (M.W.HADDOCK) (04/15/88)
I agree with Doug that Digi-View doesn't give the user enough widgets to twiddle the picture with. The other things that bugs me are that Digi-View always STRETCHES THE HISTOGRAMS for each of the color planes (not bit planes). This "usually" results in lower contrast, doesn't it? They also do no provide a library to access the little "white box". This makes it almost impossible to have my software capture its own images. This would permit time-lapsed photography er... video. It would be nice... -Rusty- ---- Rusty Haddock {uunet!likewise,cbosgd,rutgers!mtune}!hocpa!rusty AT&T Consumer Products Laboratories - Human Factors Laboratory Holmdel, New Joyzey 07733 (201) 834-1023
bph@ut-emx.UUCP (Butler Hine) (04/15/88)
[Line eater food...] I understand that the NewTek Video Toaster has frame grab capability built into it. Is this isolated, or can the Amiga access the video frame? Is the Toaster connected to the Amiga through the bus, or is it on one of the ports (serial or parallel)? Does it, in fact, digitize a full color frame in 1/60 sec? How many bits/pixel does it use internally? Enquiring minds want to know. Butler Hine Gabba, Dept. of Astronomy, University of Texas at Austin Gabba, Austin, Tx 78712 (512) 471-4419 Hey! {allegra,ihnp4}!{ut-sally,noao}!utastro!bph (internet) bph@astro.AS.UTEXAS.EDU (bitnet) bph%astro.as.utexas.edu@CUNYVM.CUNY.EDU -- Butler Hine Gabba, Dept. of Astronomy, University of Texas at Austin Gabba, Austin, Tx 78712 (512) 471-4419 Hey! {allegra,ihnp4}!{ut-sally,noao}!utastro!bph (internet) bph@astro.AS.UTEXAS.EDU (bitnet) bph%astro.as.utexas.edu@CUNYVM.CUNY.EDU