doug_rands_merritt@cup.portal.com (04/23/88)
Speaking of painful subjects that get asked over, and over, and over
again...I seem to have mislaid all of the old articles that discuss
gamma correction. Sigh. Anyone care to forward something on the topic
to me?
And since there's so much disagreement about creating *.newusers,
what about the idea of just posting "Answers to commonly asked questions"
once per month to *this* group??? (or twice monthly, given aging???)
Thanks,
Doug
Doug Merritt ucbvax!sun.com!cup.portal.com!doug-merritt
or ucbvax!eris!doug (doug@eris.berkeley.edu)
or ucbvax!unisoft!certes!dougdlou@dino.ucsd.edu (Dennis Lou) (08/26/90)
I recently read an old Macworld article on scanners. Where I work, we have a scanner with rather limited software with it. In this article, however, I read about contrast enhancement using something called Gamma Correction. I'd like to try it for the Sunraster and TIFF files we have at work. I know the basic principle of remapping the intensities, but I don't know how to compute the curve. What I'd like to know is what's the formula for Gamma Correction? Does anyone have any info they can tell me about it? THanks in advance. Please E=mail responses please! -- Dennis Lou Disclaimer: I don't use lame disks. dlou@dino.ucsd.edu "But Yossarian, what if everyone thought that way?" [backbone]!ucsd!dino!dlou "Then I'd be crazy to think any other way!"
sg04@harvey.gte.com (Steven Gutfreund) (02/22/91)
Foley & Van Dam tells me how to do gamma correction for b/w images, how do I do it for rbg? -- -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= Yechezkal Shimon Gutfreund sgutfreund@gte.com GTE Laboratories, Waltham MA harvard!bunny!sgutfreund -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
srnelson@nelsun.Eng.Sun.COM (Scott R. Nelson) (02/25/91)
From article <10652@bunny.GTE.COM>, by sg04@harvey.gte.com (Steven Gutfreund): > Foley & Van Dam tells me how to do gamma correction for b/w images, > how do I do it for rbg? You do the same thing three times, once for red, once for green and once for blue. Really. Or you can do what most graphics hardware manufacturers do and set the gamma correction value to be the same for red, green and blue. If the CRT is of good quality and is adjusted properly the gamma values for each of the three color channels will be the same. A grey-scale ramp (bars of color increasing linearly across the screen) where each intensity has equal values of red, green and blue will appear as a uniform grey with no colored spots. For those of you that don't have Foley & Van Dam or that have a hard time converting mathematic notation to code, here is a very short C program that prints a gamma correction table: /* * gamma.c * * Print a gamma table. Compile on a unix machine as follows: * cc -o gamma gamma.c -lm */ #include <math.h> #define CLUT_GAMMA 2.5 /* Gamma setting, typically 2.0 to 2.5 */ main() { int i; /* The CLUT entry number */ int g; /* The computed gamma value */ for (i = 0; i < 256; i++) { g = (int)(255.0 * pow((double) i / 255.0, 1.0 / CLUT_GAMMA)); printf("%d: %d\n", i, g); } } --- Scott R. Nelson srnelson@eng.sun.com Sun Microsystems "Proofread carefully to see if you any words out."
srnelson@nelsun.Eng.Sun.COM (Scott R. Nelson) (05/28/91)
From article <1991May27.135349.5072@vax5.cit.cornell.edu>, by wuly@vax5.cit.cornell.edu: > With regard to the "too dark" radiosity results discussion: > > After talking with a grad student in computer graphics here at cornell: > (1) everybody using radiosity algorithms has this problem. > (2) a good way to assign colors to radiosity results is to do it > linearly, and then essentially gamma-correct the resulting image. > > I have found that gamma correction does wonders for me, and my images are > darker than most due to inaccurate (too low) energy transfers (it made > calculations very fast, though). For those of you using more accurate > routines, a gamma correction style transformation during color assignment > should be helpful. > > For my case, a gamma correction of 1.8 looks nice, but of course that depends > on the monitor and the image. When is the graphics community going to catch on to the need for gamma correction in ALL shaded images? Graphics calculations need to be performed in linear space to simulate the physical mixing of light. All CRT electron guns are nonlinear. You need to use gamma correction to map these linear intensity values in the frame buffer to the nonlinear values displayed on the CRT. Gamma correction does this. The broadcast television industry has settled on a standard gamma value of 2.222222 (1.0/0.45). This has been build into television sets for decades. This value happens to look correct on all properly adjusted monitors that I have seen. Also, since gamma correction is attempting to correct for the nonlinearity of the CRT electron guns, and since the three electron guns in a color CRT are of the same design, the gamma values for the three channels should be the same. If you are computing an image using radiosity, ray tracing, antialiased lines or even gouraud shading, you need to use a gamma corrected color lookup table to properly view your image. This has been posted before, but here is a C code fragment to print a gamma correction ramp for an 8-bit lookup table: /* * gamma.c * * Print a gamma correction table. * * cc -o gamma gamma.c -lm */ #include <math.h> #define clut_gamma 2.222222 main() { int i; for (i = 0; i < 256; i++) { if (i % 16 == 0) printf("\n"); printf("%d: %d\n", i, (int) (255.0 * pow((double) i / 255.0, 1.0 / clut_gamma))); } } --- Scott R. Nelson srnelson@eng.sun.com Sun Microsystems "Proofread carefully to see if you any words out."
billd@fps.com (Bill Davidson) (05/29/91)
In article <14070@exodus.Eng.Sun.COM> srnelson@nelsun.Eng.Sun.COM (Scott R. Nelson) writes: >The broadcast television industry has settled on a standard gamma value of >2.222222 (1.0/0.45). This has been build into television sets for >decades. This value happens to look correct on all properly adjusted >monitors that I have seen. Huh? It seems like every color monitor I come across has a different gamma value and some are extreme (I know of one that needs about 3.2). Other than that, I agree with your post. Manufacturers of RGB monitors don't seem to be able to agree on anything. --Bill
bcorrie@csr (Brian Corrie) (05/29/91)
wuly@vax5.cit.cornell.edu writes: [Stuff about Radiosity problem deleted] >I have found that gamma correction does wonders for me, and my images are >darker than most due to inaccurate (too low) energy transfers (it made >calculations very fast, though). For those of you using more accurate >routines, a gamma correction style transformation during color assignment >should be helpful. Kind of a different but the same question.... I have been using SUN workstations to display generated images over the last couple of years. We recently got a NeXTstation Color. I convert the 24 bit SUN files to tiff files and display them on the NeXT, and the images are a lot darker than the images as they are displayed on the SUN. The difference is quite significant (to the point that the image isn't really worth looking at). My question is, is this a problem that could be solved by gamma correction, or is it likely a result of the conversion process? Anybody? You may have guessed that I am not an expert in Gamma correction. I always thought it was a detail for those who are perfectionists, and want an image to look identical on every monitor. Is it possible that not doing gamma correction on an image could make this much difference? Note: Some of the images are much worse than others, but all are definitely darker on the NeXTstation color than they are on the SUNs. >For my case, a gamma correction of 1.8 looks nice, but of course that depends >on the monitor and the image. >wuly@vax5.cit.cornell.edu Thanks for any help you can give me, B P.S. Is this a question for comp.graphics or comp.graphics.research? To me, it isn't research, nor is it an inquiry about research, its an enquiry about a fact. Therefore, it belongs in comp.graphics. I hope the experts don't stop reading comp.graphics so these questions will still be answered. -- Brian Corrie (bcorrie@csr.uvic.ca) Under the most rigorously controlled conditions of pressure, temperature, volume, humidity and other variables, the organism will do as it damn well pleases. Sounds like some of the code I have written...... 8-)
osmoviita@cc.helsinki.fi (05/30/91)
In article <14070@exodus.Eng.Sun.COM>, srnelson@nelsun.Eng.Sun.COM (Scott R. Nelson) writes: > > Also, since gamma correction is attempting to correct for the nonlinearity > of the CRT electron guns, and since the three electron guns in a color CRT > are of the same design, the gamma values for the three channels should be > the same. At least the people who sell monitor calibrator systems say that gamma values are different for different values -- phosphors age differently and perhaps there is some other reasons. > If you are computing an image using radiosity, ray tracing, antialiased > lines or even gouraud shading, you need to use a gamma corrected color > lookup table to properly view your image. > > This has been posted before, but here is a C code fragment to print a > gamma correction ramp for an 8-bit lookup table: > > /* > * gamma.c > * > * Print a gamma correction table. > * > * cc -o gamma gamma.c -lm > */ > > #include <math.h> > #define clut_gamma 2.222222 > > main() > { > int i; > > for (i = 0; i < 256; i++) { > if (i % 16 == 0) > printf("\n"); > printf("%d: %d\n", i, > (int) (255.0 * pow((double) i / 255.0, 1.0 / clut_gamma))); > } > } > It is not sufficient for every monitor to define gamma correction table by so simple function. And gamma values vary much between monitors and adjustments. So if you need accurate images you must adjust your monitor, measure its transfer function and calculate the correction table from your measurements. And remember to fix contrast and brightness controls unmovable and do all measuring and viewings in dark room :-) > --- > > Scott R. Nelson srnelson@eng.sun.com > Sun Microsystems > > "Proofread carefully to see if you any words out." Kari Osmoviita osmoviita@cc.helsinki.fi University of Helsinki
pierce@radius.com (Pierce T. Wetter III) (05/30/91)
>When is the graphics community going to catch on to the need for gamma >correction in ALL shaded images? >The broadcast television industry has settled on a standard gamma value of >2.222222 (1.0/0.45). This has been build into television sets for >decades. This value happens to look correct on all properly adjusted >monitors that I have seen. What does properly adjusted mean? SOAPBOX ON Television "as broadcast" is broadcast for viewing on a screen with a gamma of 2.22. Your average run of the mill TV screens have a gamma of 2.6 to 2.8. The value of 2.22 is chosen for its similarities to human vision as well as the added benifit of only correcting for gamma once at the broadcast station. By choosing 2.22, when quantizing to 8 bits, or 6 Mhz, the maximum real information bandwidth is maintained. (Humans can see 1% differences, so a perfectly perceptual space would be logorithmic, but I digress.). If you really want to adjust your image for best display, you should adjust it for a gamma correction of 2.6 to 2.8 depending on your monitor. If you are sending this out to video tape, then by all means use 2.222, since you'll at least be conistent with everyone else. If you want the best possible display, you will have to take samples of every possible value of red,green and blue, and make a conversion table. Gamma is only the first order effect, in actual monitors all sorts of other non-linearities creep in (like the brightness and contrast knobs) that require seperate treatement. >Also, since gamma correction is attempting to correct for the nonlinearity >of the CRT electron guns, and since the three electron guns in a color CRT >are of the same design, the gamma values for the three channels should be >the same. Nope, there three different guns, and the monitor has about 3 pots for each gun that can be tweaked + the brightness and contrast knobs. If you're going to approximate the brightness/voltage curve with one parameter, you'll find its different for each gun. SOAPBOX OFF Pierce -- My postings are my opinions, and my opinions are my own not that of my employer. You can get me at radius!pierce@apple.com. (Wha'ja want? Some cute signature file? Hah! I have real work to do.
pierce@radius.com (Pierce T. Wetter III) (05/30/91)
billd@fps.com (Bill Davidson) writes: >In article <14070@exodus.Eng.Sun.COM> srnelson@nelsun.Eng.Sun.COM (Scott R. Nelson) writes: >>The broadcast television industry has settled on a standard gamma value of >>2.222222 (1.0/0.45). This has been build into television sets for >>decades. This value happens to look correct on all properly adjusted >>monitors that I have seen. >Huh? It seems like every color monitor I come across has a different >gamma value and some are extreme (I know of one that needs about 3.2). >Other than that, I agree with your post. Manufacturers of RGB monitors >don't seem to be able to agree on anything. Actually, the one good thing to come out of the HDTV stuff is that they all managed to agree on one phosphor set: CCIR-Rec709, and they all pledged to use it. Of course, different coatings change the perceived color... Pierce -- My postings are my opinions, and my opinions are my own not that of my employer. You can get me at radius!pierce@apple.com. (Wha'ja want? Some cute signature file? Hah! I have real work to do.
turk@Apple.COM (Ken "Turk" Turkowski) (05/31/91)
bcorrie@csr (Brian Corrie) writes: >P.S. Is this a question for comp.graphics or comp.graphics.research? To me, >it isn't research, nor is it an inquiry about research, its an enquiry about >a fact. Therefore, it belongs in comp.graphics. I hope the experts don't stop >reading comp.graphics so these questions will still be answered. Don't assume that graphics researcher will stop reading comp.graphics just because there is a comp.graphics.research group; they'll just read the research group first. -- Ken Turkowski @ Apple Computer, Inc., Cupertino, CA Internet: turk@apple.com Applelink: TURK UUCP: sun!apple!turk
dave@imax.com (Dave Martindale) (06/06/91)
In article <14070@exodus.Eng.Sun.COM> srnelson@nelsun.Eng.Sun.COM (Scott R. Nelson) writes: > >The broadcast television industry has settled on a standard gamma value of >2.222222 (1.0/0.45). This has been build into television sets for >decades. This value happens to look correct on all properly adjusted >monitors that I have seen. Well, actually, NTSC specifies a monitor gamma of 2.2 exactly, giving a camera (or lookup table, since your frame buffer mimics a camera) of 0.4545. For practical purposes, 0.45, but it is the 2.2 number that is defined, not 0.45. This was done with full knowledge that the "average" receiver has a gamma of about 2.8, because they wanted the image on screen to have a gamma higher than is "realistic" because people prefer that when looking at images in a room with dimmed lighting. The result of gamma correction for 2.2 followed by display on a monitor with a real gamma of 2.8 is an image with a gamma of 2.8/2.2 = 1.27 relative to the original scene. PAL specifies a monitor gamma of 2.8 instead of 2.2. This should give more accurate tone reproduction on sets with a real gamma of 2.8. However I have heard at least one report that European broadcasters adjust cameras for a gamma of 0.45 anyway, not 0.36, because they want the same contrast increase that the NTSC standard gives. (Is there anyone out there who works in a European television studio? Are cameras set up for a gamma of 0.45 or 0.36, or something else entirely?) The HDTV production standard specifies a more complex "opto-electronic transfer characteristic" that is basically gamma correction with a factor of 0.45 in bright areas, spliced to a linear function at low brightnesses. They did this to have a characteristic that camera gamma-correctors could adhere tightly to; the NTSC and PAL gamma-correction functions are not physically realizable for dark areas of the picture. So, the "gamma correction" function you should use is standardized, but the standard you should use depends on the television system you are encoding your picture for. On the other hand, if you just want to display your picture on a monitor, you should gamma-correct for that particular monitor. If you want to accurately reproduce tones, you should fully correct for the monitor gamma, to give an on-screen picture with a gamma of 1.0 relative to the calculated intensities. If you want to see what the image will look like on TV, though, you might want to use a lookup table that gives the on-screen picture a gamma of about 1.25 relative to the calculated intensities, since NTSC encoding will do that. And if you want to see what it will look like when recorded on film and projected in a theatre, you might want to adjust the on-screen picture for a gamma of 1.5, since that's what film does. Of course, as Scott points out, not doing any gamma correction at all will get you an extremely dark picture that bears little relationship to the intensities you calculated. It's better to gamma-correct using almost any value of gamma than to ignore the problem. Picking the *precise* value of gamma that is appropriate to your circumstances is less important. Dave Martindale
dave@imax.com (Dave Martindale) (06/06/91)
In article <1466@radius.com> pierce@radius.com (Pierce T. Wetter III) writes: > >>Also, since gamma correction is attempting to correct for the nonlinearity >>of the CRT electron guns, and since the three electron guns in a color CRT >>are of the same design, the gamma values for the three channels should be >>the same. > > Nope, there three different guns, and the monitor has about 3 pots for each > gun that can be tweaked + the brightness and contrast knobs. If you're > going to approximate the brightness/voltage curve with one parameter, you'll > find its different for each gun. All the discussion about gamma correction assumes that the monitor has been properly adjusted in the first place. The black level should be set so that the guns are just barely cut off with zero input voltage, and so that very dim greys are neutral in colour. The individual colour gains should be set to give white of the correct colour temperature. A grey scale that goes from black to white should remain the same colour temperature throughout the scale. If you've adjusted your monitor so it meets all these requirements, I think you'll find that this constrains the adjustment of almost every control in the monitor except "contrast". When this is done, you should find that the brightness/voltage curves for the three colours track quite accurately (if they didn't, the grey scale would show colour shifts). And that a single value for gamma pretty accurately describes how the monitor responds to voltage. If the monitor is misadjusted, then the best gamma-correction value may be different for each colour. But in that case, the transfer function of at least one of RGB is likely to look nothing like the power-law function that gamma correction models, and gamma correction isn't going to provide good compensation for this. You need something more powerful than simple gamma correction. Fixing the monitor's calibration is probably easier than fixing the model. If you want to get more accurate, you could take a bunch of brightness vs. voltage measurements and fit a curve to it - the relationship of brightness to voltage is not an exact power-law curve, and a bunch of numbers can describe it better than one. But still, the three colours should track each other pretty closely unless the monitor is misadjusted. Dave Martindale