munish@ms.uky.edu (Munish Mehra) (12/06/89)
There are many scanners that scan images into 16,32 or more gray levels . (Some very reasonable) Is it possible to use color filters / colored cellophane paper to scan an image through Red, Green and Blue Filters and then combine the RGB values to create a colored image ? Since color scanners are far more expensive than gray-scale scanners, it seems that good color images could be created for a fraction of the cost. (I think voyager did something like that to create the color images) ------------------------------------------------------------------- Anti-PC-People STOP NOW. DO NOT READ FURTHER. Can we use a $175-(DFI-HS-3000)-like, cheap scanner (which I just purchased), to do the above? There would be some programming required in combining the RGB levels to get a good scan. But with 32 levels of each R,G & B. I don't see why we couldn't get near photographic quality color images, at a fraction of the cost of color scanners. Any Help on the above would be appreciated ?
TEG@orc.olivetti.com (12/06/89)
You definately cannot do color scanning (using color filters) with a hand scanner. The problem is that hand scanners use a narrow spectrum of light (usually visible red or green) to detect the image. If you put a red filter on a hand scanner with red light source you would get a blank scan. Some photocopiers used bluish or greenish light sources so it was known to many a secretary not to try copying documents written in green or blue ink... Most of the color scanners use multiple light sources (ie. RGB) with one scanning element (and multiple passes - 1 for each color). Color filter scanning works well with b/w t.v. cameras and a frame grabber board. Many setups of this type were sold to amiga users. It takes a while to flip the "color wheel" between scans and your subject must stay still. Tom Griner Systems Administrator - Olivetti Research Center frames 2 /dev/fb uunet!wyse!decwrl!pyramid!oliveb!TEG TEG@ORC.Olivetti.Com flames 2 /dev/null pqpqpqpqpqpqpqpqpqpqpqpqpqpqpqpqpqpqpqpqpqpqpqpqpqpqpqpqpqpqpqpqpqpqpqpqpqpqpq bdbdbdbdbdbdbdbdbdbdbdbdbdbdbdbdbdbdbdbdbdbdbdbdbdbdbdbdbdbdbdbdbdbdbdbdbdbdbd
tms@hprnd.HP.COM (Thomas Skibo) (12/07/89)
/ hprnd:comp.graphics / munish@ms.uky.edu (Munish Mehra) / 2:25 pm Dec 5, 1989 / > There are many scanners that scan images into 16,32 or more gray levels . > (Some very reasonable) > Is it possible to use color filters / colored cellophane paper to > scan an image through Red, Green and Blue Filters and then combine > the RGB values to create a colored image ? I tried a similar approach with a B/W video camera and did not achieve very good results. The problem was that the video camera was not equally sensitive to all visible colors. The three filtered images were not true representations of the red, green, and blue components. Some problems you may run into with the scanner: 1) The light source in the scanner is not bright enough to go through the filter paper and the scanner's dark/light adjustment cannot compensate. 2) The light source in the scanner doesn't cover the visible color spectrum very well thus making it really tough to capture one or more of the primary color components. 3) The scanner is not equally sensitive to all the colors. Thomas Skibo Roseville Networks Div. Hewlett-Packard tms@hprnd.hp.com
chris@com2serv.C2S.MN.ORG (Chris Johnson) (12/08/89)
In article <13428@s.ms.uky.edu> munish@ms.uky.edu (Munish Mehra) writes: > < stuff deleted > >to get a good scan. But with 32 levels of each R,G & B. I don't see >why we couldn't get near photographic quality color images, at a >fraction of the cost of color scanners. Maybe this goes without saying in this newsgroup, but 24 bits of color value for RGB might be called near photographic quality, as far as color rendition goes, but 5 bits (32 levels) probably is not very close. And that's only color rendition. Even a 400dpi image is not remotely close to photographic quality in resolution. I don't have any exact figures at hand, but I would suspect something like Kodachrome could do well over 10,000dpi. Of course, an ISO 1600 film might be only 600dpi! :-) Chris Johnson UUCP: chris@c2s.mn.org Com Squared Systems, Inc., USA ATT: +1 612/452-9522
darcy@druid.uucp (D'Arcy J.M. Cain) (12/08/89)
In article <3184@com50.C2S.MN.ORG> chris@com2serv.c2s.mn.org (Chris Johnson) writes: >In article <13428@s.ms.uky.edu> munish@ms.uky.edu (Munish Mehra) writes: >>to get a good scan. But with 32 levels of each R,G & B. I don't see > >Maybe this goes without saying in this newsgroup, but 24 bits of color >value for RGB might be called near photographic quality, as far as color >rendition goes, but 5 bits (32 levels) probably is not very close. In fact it isn't quite that bad. Munish mentions 32 levels each so the comparison is 5 to 8 or 15 to 24, not 5 to 24. This of course isn't phoygraphic quality but some of the images done on 15 bit (usually 16 bit with one bit for overlay) display boards can be quite impressive depending on the rendering software. -- D'Arcy J.M. Cain (darcy@druid) | "You mean druid wasn't taken yet???" D'Arcy Cain Consulting | - Everybody - West Hill, Ontario, Canada | No disclaimers. I agree with me |
dave@imax.com (Dave Martindale) (12/13/89)
In article <3184@com50.C2S.MN.ORG> chris@com2serv.c2s.mn.org (Chris Johnson) writes: > >Maybe this goes without saying in this newsgroup, but 24 bits of color >value for RGB might be called near photographic quality, as far as color >rendition goes, but 5 bits (32 levels) probably is not very close. Well, actually, even 8 bits isn't enough to handle the brightness range that film does (particularly negative film) without getting "banding" effects due to quantization error. 12 bits per component is about right. (Actually, 8 bits would be OK if there was a linear-to-log converter in front of the A/D, but I haven't seen a commercial digitizer that does this. I wonder why?) >And that's only color rendition. Even a 400dpi image is not remotely >close to photographic quality in resolution. I don't have any exact >figures at hand, but I would suspect something like Kodachrome could >do well over 10,000dpi. Of course, an ISO 1600 film might be only >600dpi! :-) At high contrast, most colour films can resolve 100 or 125 line pairs (at 2 pixels/line pair) per mm. 100 lp/mm is about 5000 pixels/inch. There are slow B&W films (Tech Pan) that have 2 or 3 times this resolution when exposed and developed appropriately. To be fair, this is the upper resolution limit of film, where it can just barely resolve something that had a 1000:1 contrast ratio in the original. Most film will resolve only about 30 lp/mm at low contrast. So, if an electronic image is to look "as sharp" as a film one, the number of pixels you need depends on whether the image has sharp, high-contrast edges or not. Another factor is that, with movies, a series of frames projected at 24 FPS looks sharper than any single frame from the series does - the noise due to film grain averages out. Projecting at 60 FPS (e.g. Showscan) gets a further gain this way. Electronic images won't gain in the same way. When comparing film to electronic imagery, you also have to take into account the magnification you will subject the film image to. For example, a 35mm slide is 36mm wide. At 30 lp/mm, that's 2160 pixels wide. If you enlarge that to 7.2 inches wide on a print, it's only 300 dpi. (Well, except that the high-contrast features have an effective resolution of 1000 dpi). But if 35mm doesn't give you a sharp enough image at the enlargement you want, you just use a bigger piece of film. That's what's so nice about film - its flexibility. If you want more resolution, you just use more film area and change the lens focal length. With electronic imagery, doubling the linear resolution (and quadrupling the pixel count) is a horrendous problem. For some real-world comparisons: A reasonable figure for the resolution of film when projected on screen in a theatre is perhaps 36 lp/mm. This is what the audience can actually see, and includes losses in photography, printing, and projection. Then normal 35mm 4 perf film, as seen in your local theatre, gives an image that contains about 1447 x 1094 "resolvable pixels" (assuming that the resolution at the edge of the screen is the same as the centre - it isn't). 16mm film gives about 690 x 505 pixels. 35mm 8-perf, the format used by still cameras, gives 2600 x 1730 pixels 70mm 5 perf (theatrical "70 mm" format and Showscan) is capable of 3500 x 1590 "resolvable pixels". And IMAX (70mm 15 perf) gives 5010 x 3490 pixels. NTSC video is capable of at most 370 x 350 (remember, this is what the viewer can resolve, not what your frame buffer stores for NTSC output). (And NTSC's colour resolution is far worse than its luminance resolution, by a factor of 3 to 9.) HDTV resolution has been measured at about 1030 x 1030 "resolvable pixels". In other words, HDTV has finally surpassed 16mm film in quality, but does not equal even moderate-quality 35mm. For a flatbed scanner, you would have to put a resolution test target on it and see what the finest pattern it can actually resolve is. Although a scanner may be 300 DPI, it is not likely to actually resolve 150 lp/inch, the theoretical limit, for a number of reasons. My guess would be that you might actually see 100 lp/inch, for about 2000 "resolvable pixels" across a 10-inch image width. So a 300 dpi 8x10 inch flatbed scanner might have better "information capacity" than 35mm movie film, but less than a 35mm still-camera slide.
raveling@isi.edu (Paul Raveling) (12/16/89)
In article <3184@com50.C2S.MN.ORG>, chris@com2serv.C2S.MN.ORG (Chris Johnson) writes: > > Maybe this goes without saying in this newsgroup, but 24 bits of color > value for RGB might be called near photographic quality, as far as color > rendition goes, but 5 bits (32 levels) probably is not very close. It depends a lot on the image. For most images 5 or 6 bits is about the threshold where a casual viewer won't notice the difference when comparing them with any higher number of levels. On the extremes, there are a few images where this is true at 4 bits, and relatively more that need about 8 bits or more. I can also believe reports that some images improve when viewed with 30 or 36 bits rather than 24, because human perception is highly sensitive to differences in color rather than absolute color. I suspect that it might be possible to quantize images to perhaps 10-12 bits and get the same effect; we're limited a bit by hardware what usually shows either 24-bit unquantized color or at most 8-bit quantized color. > And that's only color rendition. Even a 400dpi image is not remotely > close to photographic quality in resolution. Most people swear my 100 dpi monitor produces photographic quality resolution. In truth it doesn't, but usually the mind of the viewer does. In fact it's fun to occasionally use a magnifier window to demonstrate to people that some of the detail they see doesn't exist in the image, but rather is synthesized in their brain based on some subtle cues that do exist. ---------------- Paul Raveling Raveling@isi.edu
littauer@uts.amdahl.com (Tom Littauer) (12/21/89)
In article <2300002@hprnd.HP.COM> tms@hprnd.HP.COM (Thomas Skibo) writes: > >/ hprnd:comp.graphics / munish@ms.uky.edu (Munish Mehra) / 2:25 pm Dec 5, 1989 / >> There are many scanners that scan images into 16,32 or more gray levels . >> (Some very reasonable) >> Is it possible to use color filters / colored cellophane paper to >> scan an image through Red, Green and Blue Filters and then combine >> the RGB values to create a colored image ? > >I tried a similar approach with a B/W video camera and did not achieve very >good results. The problem was that the video camera was not equally sensitive >to all visible colors. The three filtered images were not true representations >of the red, green, and blue components. This is true, but you can scale the individual components. Other problems I've run into using RS-170 cameras: Many are sensitive to IR, and most RGB filters pass IR. Get an IR-blocking filter for best results. Be sure to include samples of pure black and pure white (as you define them) in each frame you digitize for scaling purposes. Many cameras do some form of AGC even if you turn off all the widgets marked AGC. Good Luck, Tom Littauer -- UUCP: littauer@amdahl.amdahl.com or: {sun,decwrl,hplabs,pyramid,ames,uunet}!amdahl!littauer DDD: (408) 737-5056 USPS: Amdahl Corp. M/S 278, 1250 E. Arques Av, Sunnyvale, CA 94086 I'll tell you when I'm giving you the party line. The rest of the time it's my very own ravings (accept no substitutes).
tms@hprnd.hp.com@canremote.uucp (tms@hprnd.HP.COM) (12/21/89)
From: tms@hprnd.HP.COM (Thomas Skibo) Orga: HP Roseville Networks Division / hprnd:comp.graphics / munish@ms.uky.edu (Munish Mehra) / 2:25 pm Dec 5, 1989 / > There are many scanners that scan images into 16,32 or more gray levels . > (Some very reasonable) > Is it possible to use color filters / colored cellophane paper to > scan an image through Red, Green and Blue Filters and then combine > the RGB values to create a colored image ? I tried a similar approach with a B/W video camera and did not achieve very good results. The problem was that the video camera was not equally sensitive to all visible colors. The three filtered images were not true representations of the red, green, and blue components. Some problems you may run into with the scanner: 1) The light source in the scanner is not bright enough to go through the filter paper and the scanner's dark/light adjustment cannot compensate. 2) The light source in the scanner doesn't cover the visible color spectrum very well thus making it really tough to capture one or more of the primary color components. 3) The scanner is not equally sensitive to all the colors. Thomas Skibo Roseville Networks Div. Hewlett-Packard tms@hprnd.hp.com --- * Via MaSNet/HST96/HST144/V32 - UN Graphics * Via Usenet Newsgroup comp.graphics
phorgan@cup.portal.com (Patrick John Horgan) (01/04/90)
Just this technique is used by the Digi-View scanner by NewTek. This scanner (for the Amiga), uses black and white video cameras with a color wheel...their software even controls the motor to rotate the wheel...Their results are excellent! Patrick Horgan phorgan@cup.portal.com
phorgan@cup.portal.com (Patrick John Horgan) (01/05/90)
Anyone know where I can get specs on the C language bindings for GKS? (3d GKS too:) EMAIL me @ phorgan@cup.portal.com Thanks
awpaeth@watcgl.waterloo.edu (Alan Wm Paeth) (01/10/90)
In <5442@udccvax1.acs.udel.EDU> cygnus@vax1.acs.udel.EDU (Marc Cygnus) writes: >In <1941@hydra.riacs.edu> hitchner@same (Lew Hitchner) writes: >> The filters you should buy are called wratten filters... > >Quite correct, Lew. Perhaps the following information would be of use... > >Kodak Wratten gelatine filters (you can get glass, but I chose the gels) >No. 29 (red) #149 5621 \ >No. 61 (green) #149 5894 +-- Kodak catalogue #s for 75mm x 75mm >No. 47 (blue) #149 5787 / NOTE: most often the 25 Red and 47B Blue are employed for photographic color separation work -- the set you choose may be different as a function of the underlying CCD array (or film, photomultiplier tube, etc.) For what its worth, my reference on commercial printing freely substitutes a 29 red in some cases but cautions against use of the blue "straight 47" as an alternate to the 47B. If someone has a copy of the Kodak Wratten filter book on hand perhaps they can furnish numbers on the cutoff points? It would be nice to know the whole truth. /Alan Paeth Computer Graphics Laboratory University of Waterloo
dave@imax.com (Dave Martindale) (01/11/90)
In article <12887@watcgl.waterloo.edu> awpaeth@watcgl.waterloo.edu (Alan Wm Paeth) writes: > >If someone has a copy of the Kodak Wratten filter book on hand perhaps they can >furnish numbers on the cutoff points? It would be nice to know the whole truth. Here are the figures for various interesting filters. I've shown the approximate wavelengths where transmission falls to 10% and 1% (OD 1.0 and 2.0). Wavelengths are in nanometers: Wratten # 1% 10% peak 10% 1% peak transmission (%) 25 580 590 89 26 590 600 89 29 600 610 90 47 380 400 440 500 520 50 47B 380 400 430 470 490 50 48 400 420 460 490 510 33 58 480 500 530 580 600 54 61 480 500 520 570 600 40 Dave Martindale