RE: filmscanners: creating correction curves from scanned calibration chart?
Ned writes ... From: Todd Flashner [EMAIL PROTECTED] However, for a neutral gray they (RGB) SHOULD all be the same number. no, total rubbish ... the profile defines the relationship between the value and the actual colour output Quite right, but we're talking about typical PS working spaces ... whereby 0,0,0 is absolute black, 255,255,255 is absolute white ... and neutral grays all are defined by R=G=B. They may not be realistic, but we then depend on device spaces for going to, or coming from, working space to reality. shAf :o)
Re: filmscanners: creating correction curves from scanned calibration chart?
Hey there, There is a problem here, and that is your assumption that the 18% grey would be at RGB 117/117/117. RGB is a totally device specific colour space and so you are completely at the mercy of the video card and monitor and resolution. The correct value for mid grey would be... whatever the RGB value you have measured off your monitor that matches mid grey. if you change your monitor/card/resolution you have to re-measure. If you are under different lighting - you have to re-measure. Basically - the only way to do this successfully is to get one of those monitor measuring devices and profile your monitor, and calibrating your workflow. I worked on the front end to Agfa Colourtune a while ago and thus worked with the colour scientists a lot. I remember one guy saying to me that somebody had published the RGB values for mid-grey and how this was crazy. Stuck in my head. Regards Ned From: michael shaffer [EMAIL PROTECTED] Reply-To: [EMAIL PROTECTED] To: film scanner list [EMAIL PROTECTED] Subject: filmscanners: creating correction curves from scanned calibration chart? Date: Fri, 9 Nov 2001 19:22:37 -0330 I have a scan of a photographed color chart, and I also have the logD reflectance values for the patchs. I want to create Photoshop curves for correcting the scan ... for example, assuming gamma=2.2 working space, I calculate a value of 117 for 18% reflectance: gray_value == (0.18^(1/2.2))*255 = 117 . Similarly, logD (18%) = .745 implies (1/(10^logD))^(1/2.2)*255 = 117 (sorry ... it looks much better in a spreadsheet) so ... if my scanned RGB values for 18% gray are 115/118/116, I should create a curve for transforming these to 117/117/117. If correct, I have 2 questions: (1) Although I may be satisfied with neutralizing gray patchs, if I have similar logD numbers for the colored patchs, can I apply the same formula to each of the RGB channels? (2) What is the best way to create the curves? For example, I would like to use the target for the curve's gray eyedropper to change the curve such that 115/118/116 becomes 117/117/117, ... and for a different gray, 155/159/154 becomes 157/157/157 ... but the gray eyedropper doesn't seem to work this way(?) tia ... shAf :o) _ Get your FREE download of MSN Explorer at http://explorer.msn.com/intl.asp
RE: filmscanners: creating correction curves from scanned calibration chart?
Ned writes ... There is a problem here, and that is your assumption that the 18% grey would be at RGB 117/117/117. ... The correct value for mid grey would be... whatever the RGB value you have measured off your monitor that matches mid grey. ... ... Basically - the only way to do this successfully is to get one of those monitor measuring devices and profile your monitor, and calibrating your workflow. Because my system is calibrated, I am not assuming. If my monitor is truely gamma calibrated, and I use a gamma=2.2 working color space, then a 18% gray card's RGB value should be R=G=B=117 (see formulas below, they aren't mine). I ask you ... under what circumstances would an 18% gray card be anything else? (you may also assume I have ambient illumination under control.) However, gray balancing is one thing ... involving the color patches and color space gamut is another, and therefore the reason form my #1 question. However, I'll probably satisfy myself having gray balanced (as Bruce Fraser implies ... you're 95% there if the grays are accurate). The more important question is still #2 ... how do I do it?. shaf :o) From: michael shaffer [EMAIL PROTECTED] gray_value == (0.18^(1/2.2))*255 = 117 . Similarly, logD (18%) = .745 implies (1/(10^logD))^(1/2.2)*255 = 117 (2) What is the best way to create the curves? For example, I would like to use the target for the curve's gray eyedropper to change the curve such that 115/118/116 becomes 117/117/117, ... and for a different gray, 155/159/154 becomes 157/157/157 ... but the gray eyedropper doesn't seem to work this way(?)
Re: filmscanners: creating correction curves from scanned calibration chart?
If you're using PS, click on each the points in the image that you want to set, check the read numbers, then click, in the separate channels, on the curve at or about that point, they type in the read number in the input box and your desired output number in the output box. Do this for each point you want to set and in each channel. There may be a quicker way to set the input point on the line, maybe by double-clicking on the image spot, but I don't remember for sure. Maris - Original Message - From: michael shaffer [EMAIL PROTECTED] To: [EMAIL PROTECTED] Sent: Sunday, November 11, 2001 7:59 AM Subject: RE: filmscanners: creating correction curves from scanned calibration chart? [snippped] | However, I'll probably satisfy myself having gray balanced (as Bruce | Fraser implies ... you're 95% there if the grays are accurate). The more | important question is still #2 ... how do I do it?. | | | shaf :o) | | | From: michael shaffer [EMAIL PROTECTED] | | gray_value == (0.18^(1/2.2))*255 = 117 . Similarly, | | logD (18%) = .745 implies (1/(10^logD))^(1/2.2)*255 = 117 | | (2) What is the best way to create the curves? For example, | I would like to use the target for the curve's gray eyedropper | to change the curve such that 115/118/116 becomes 117/117/117, | ... and for a different gray, 155/159/154 becomes 157/157/157 | ... but the gray eyedropper doesn't seem to work this way(?) | | |
Re: filmscanners: creating correction curves from scanned calibration chart?
If you're using PS, click on each the points in the image that you want to set, check the read numbers, then click, in the separate channels, on the curve at or about that point, they type in the read number in the input box and your desired output number in the output box. Do this for each point you want to set and in each channel. There may be a quicker way to set the input point on the line, maybe by double-clicking on the image spot, but I don't remember for sure. I think if you control-shift-click on each point it will set those points on the individual channel curve line. -- Bob Shomler http://www.shomler.com/gallery.htm
RE: filmscanners: creating correction curves from scanned calibration chart?
your assumption being that 0,0,0 is totally black and 255,255,255 is totally white in that 2.2 gamma colour space. Would be a pretty daft colour space as you can't get either on a monitor or printer and so you end up wasting a whole bunch of values that could never be properly expressed. Huh? Would you elaborate on this please.
RE: filmscanners: creating correction curves from scanned calibration chart?
Ned writes ... From: michael shaffer [EMAIL PROTECTED] Because my system is calibrated, I am not assuming. If my monitor is truely gamma calibrated, and I use a gamma=2.2 working color space, then a 18% gray card's RGB value should be R=G=B=117aren't mine). ... your assumption being that 0,0,0 is totally black and 255,255,255 is totally white in that 2.2 gamma colour space. Would be a pretty daft colour space as you can't get either on a monitor or printer ... generate an icm profile for you monitor (i.e. with a color sensor) and scanner and ignore the rgb values. Quite independent of whether my monitor or scanner have an accurate device profile, is whether the scanner software accurately calculated the whitepoint for a given film scan. There are numerous examples for film combinations and real life whereby the scanner software is not given enough info, or assumes the wrong RGB information, and subsequently calculates a wrong whitepoint (e.g., garden scenery or miscalculating the orange mask). The result, of course, gamma space is miscalculated, as is well the neutrality of what should be gray. It is this problem I'm trying to address by photographing a color chart ... I've already taken care of the rest of my hardware. shAf :o)
RE: filmscanners: creating correction curves from scanned calibration chart?
Maris writes ... If you're using PS, click on each the points in the image that you want to set, check the read numbers, then click, ... Bob writes ... I think if you control-shift-click on each point it will set those points on the individual channel curve line. ... Control-clikking a region in the image allows me to manually set the curve ... for each RGB channel ... a bit tedious. If you can't use the gray eyedropper (and its target) to easily do this, then I'm apparently missing the point of the gray eyedropper (and its target). thanx ... shAf :o) | From: michael shaffer [EMAIL PROTECTED] | (2) What is the best way to create the curves? For example, | I would like to use the target for the curve's gray eyedropper | to change the curve such that 115/118/116 becomes 117/117/117, | ... and for a different gray, 155/159/154 becomes 157/157/157 | ... but the gray eyedropper doesn't seem to work this way(?) | | |
RE: filmscanners: creating correction curves from scanned calibration chart?
If you're using PS, click on each the points in the image that I think if you control-shift-click on each point it will set those points on the individual channel curve line. ... Control-clikking a region in the image allows me to manually set the curve ... for each RGB channel ... a bit tedious. If you can't use the gray eyedropper (and its target) to easily do this, then I'm apparently missing the point of the gray eyedropper (and its target). thanx ... shAf :o) Control AND shift together set curve points differently than control-click by itself. Control-click sets a curve point on the RGB curve; control+shift+click sets a point on each of the color channel curves. So for example, an eyedropper over a point that shows R77 G69 B66, control-clicked onto a curve, will put a point on the RGB curve at 71, while a control+shift+click will put points on the individual channel curves: 77 on red, 69 on green, 66 on blue. The points, set by control or control+shift, do not change anything. Concept is that you go to individual curves and use the click-set points to adjust to suit. This was responding to your comment There may be a quicker way to set the input point on the line, maybe by double-clicking on the image spot, but I don The grey eyedropper is different; it changes the individual channel curves. It applies a change point to each of the channel curves so that the input is the value of that channel under the eyedropper and the output point is set to 128, midpoint grey: R=G=B=128. The idea is that this control+shift click is a quicker and usually more precise than clicking on each individual channel. -- Bob Shomler http://www.shomler.com/gallery.htm
Re: filmscanners: creating correction curves from scanned calibration chart?
From: John Brownlow [EMAIL PROTECTED] this is nuts. loads of people, me included, edit by the numbers. The whole point of an ICM profile is so that the same RGB values display the same on different profiled devices. Sorry to tell you, but you are *so* wrong. RGB values mean diddly squat. They are just a measurement of the amount of current in a scanner ccd or the output voltage of a video card. The ICM profile provides a description of the transfers that are required to turn the RGB values into proper colimetric tristimulus values (i.e. XYZ). That then enables transfers between colour spaces to happen. (extreme example) You could have a scanner that mapped mid grey at 20,20,20 and a monitor that mapped it at 230,230,230. The ICM Profile describes this and allows colour management systems to convert from one colour space to another, and the RGB (device-dependant) values change accordingly, depending on which space you are in, but the device-independant colour space values (XYZ, LAB, X'Y'Z' etc) would not change. Incidentally, 18% grey means that the card REFLECTS 18% of the light. So it is a relative grey (relative to 100% reflectance of the incident light). Not to 'pure' white, whatever that may be. It all gets majorly complex when talking about scanning film, as the CCDs are only registering a value that they see between their black and white points. The first problem with doing all of this is 'has the film been processed exactly the same way as the one the profile was generated from' and 'is the exposure totally spot on? The answer to both of them is generally no :-) _ Get your FREE download of MSN Explorer at http://explorer.msn.com/intl.asp
RE: filmscanners: creating correction curves from scanned calibration chart?
jeez its late and i've had too many beers but ill give it a go. (extreme example) imagine we had a linear luminocity scale of 0-100 (0% to 100%) and these are represeted in values in a file between 0-100. The best device you have got can represent say 10%-90%. anything below 3 or above 96 is clamped at the respective values. that means your entire contrast range can only be expressed in the 80 steps between 10 and 89, thus leaving 20 values unavailable to be used. If we all agreed that the 0-100 values in the file mapped to the real values of 10% to 90%, then we have 100 steps of contrast different between the two points? Not making sense? hmm, even I'm not sure what I'm writing at the mo. This is my last mail on the subject. topic closed. further reading: http://www.inforamp.net/~poynton/notes/colour_and_gamma/ColorFAQ.html warmest regards Ned (or Ed as shAf called me, lol) From: Austin Franklin [EMAIL PROTECTED] Reply-To: [EMAIL PROTECTED] To: [EMAIL PROTECTED] Subject: RE: filmscanners: creating correction curves from scanned calibration chart? Date: Sun, 11 Nov 2001 14:54:01 -0500 your assumption being that 0,0,0 is totally black and 255,255,255 is totally white in that 2.2 gamma colour space. Would be a pretty daft colour space as you can't get either on a monitor or printer and so you end up wasting a whole bunch of values that could never be properly expressed. Huh? Would you elaborate on this please. _ Get your FREE download of MSN Explorer at http://explorer.msn.com/intl.asp
RE: filmscanners: creating correction curves from scanned calibration chart?
RGB values mean diddly squat. They are just a measurement of the amount of current in a scanner ccd Only if it is a raw scan. A typical scan has the data setpointed and tonally adjusted. That means that 0 IS the darkest value in the resultant file, and 255 IS the lightest value in the file, if we are talking about 8 bit files that is. or the output voltage of a video card. Yes, bit as above, it's NOT raw data... I really think you're missing some basic understanding here... (extreme example) You could have a scanner that mapped mid grey at 20,20,20 Huh? What does that mean. Scanners don't just map mid gray to some value. I really think you don't know how a scanner really works. I suggest you search the archives for some more info, or see Wayne Fulton's www.scantips.com for more info on how scanners work. It all gets majorly complex when talking about scanning film, as the CCDs are only registering a value that they see between their black and white points. That is not right. The CCD does not have a black and white point. That gets set in software, or in an analog front end (but not in any regular scanner). The first problem with doing all of this is 'has the film been processed exactly the same way as the one the profile was generated from' and 'is the exposure totally spot on? The answer to both of them is generally no :-) Well, you got that right. That is why film profiling is really not very useful.
Re: filmscanners: creating correction curves from scanned calibration chart?
- Original Message - From: Ned Nurk [EMAIL PROTECTED] To: [EMAIL PROTECTED] Sent: Sunday, November 11, 2001 11:36 PM Subject: Re: filmscanners: creating correction curves from scanned calibration chart? From: John Brownlow [EMAIL PROTECTED] this is nuts. loads of people, me included, edit by the numbers. The whole point of an ICM profile is so that the same RGB values display the same on different profiled devices. Sorry to tell you, but you are *so* wrong. RGB values mean diddly squat. They are just a measurement of the amount of current in a scanner ccd or the output voltage of a video card. The ICM profile provides a description of the transfers that are required to turn the RGB values into proper colimetric tristimulus values (i.e. XYZ). That then enables transfers between colour spaces to happen. (extreme example) You could have a scanner that mapped mid grey at 20,20,20 and a monitor that mapped it at 230,230,230. The ICM Profile describes this and allows colour management systems to convert from one colour space to another, and the RGB (device-dependant) values change accordingly, depending on which space you are in, but the device-independant colour space values (XYZ, LAB, X'Y'Z' etc) would not change. Incidentally, 18% grey means that the card REFLECTS 18% of the light. So it is a relative grey (relative to 100% reflectance of the incident light). Not to 'pure' white, whatever that may be. It all gets majorly complex when talking about scanning film, as the CCDs are only registering a value that they see between their black and white points. The first problem with doing all of this is 'has the film been processed exactly the same way as the one the profile was generated from' and 'is the exposure totally spot on? The answer to both of them is generally no :-) _ Get your FREE download of MSN Explorer at http://explorer.msn.com/intl.asp Surely for each colour space there is a particular value for each representable colour of that colour space. When you look at the values in PS they have a precise translation to a particular colour. Quite what their equivalent values would be for your monitor, printer or scanner is irrelevant as that is the whole point of using an ICM profile for each device - you don't need to know. You get your data from the scanner and the profile is used to convert to your working space. The data in you working space is converted by the monitor profile so that you see the correct colour for the correct RGB values of the working space. Similarly the printer profile is used to covert from the working space so that the correct colour is produced on your printer. Sadly in practice profile conversions can never be entirely accurate as each device can only represent a particular set of colours. Hence you should use a colour space with a wide gamut for editting. This gives you the best chance of least inaccurate colour reproduction on each device. Steve
