I have been thinking about using a digital camera to see how far from D65 my projector is currently calibrated to. Here is my idea:

I set the white balance on my camera by using a white paper (or 18% gray card) in daylight (approx D65). Next step is to go into my theatre room (as good as pitch black in there) and then take a picture of my projector screen when it is displaying a 100 IRE calibration image (or anything else that the projector believes to be white). My thought is that if I have a blue or red color cast that means that the color temperature is off.

This idea is very simplistic and I was just wondering if I am missing something obvious or maybe I am just not grasping the concept behind color temperature and how to measure it.

Appreciate any feedback.

I have been thinking about using a digital camera to see how far from D65 my projector is currently calibrated to. Here is my idea:

I set the white balance on my camera by using a white paper (or 18% gray card) in daylight (approx D65). Next step is to go into my theatre room (as good as pitch black in there) and then take a picture of my projector screen when it is displaying a 100 IRE calibration image (or anything else that the projector believes to be white). My thought is that if I have a blue or red color cast that means that the color temperature is off.

This idea is very simplistic and I was just wondering if I am missing something obvious or maybe I am just not grasping the concept behind color temperature and how to measure it.

Appreciate any feedback.Offhand, I see two specific issues with this approach:
1) You need to be able to force the camera to NOT auto adjust white balance or color - most cameras these days auto adjust everything.
2) "Daylight" basically has a large variation in color temp (time of day, day of year, longitude/latitude, conditions of the sky (overcast, direct sun, etc.).

How about holding up a card with a known color (eg your grey card) in front of the 100 IRE window and then taking the picture. With the card and the 100IRE on the same picture you could do some exploring?

Problem is, even if this works, is it practical to take a picture, evaluate it, make adjustments, take a picture, evaluate it, etc...?

Mark

Thanks for your comments.

The idea would be to use a camcorder to look at the screen and then have that feeding into a computer where my custom software would figure out how far from the right color temperature the image and then have the software control via the rs-232 interface the projector settings. I am just in the exploratory phase of this application right now hence the usage of a digital camera to validate/invalidate my ideas.

I thought about holding the grey card next to the 100 IRE image so they both get the same illumination. However, I still can't figure out if that will work since wouldn't that be saying that the projector's color temperature is correct? The idea I had was to make sure when the projector outputs a white image it looks the D65 white should look like. I am assuming that when they master movies and other source material that there is an assumption that the colors have been adjusted to match D65.

Don't forget, that the color temp needs to be correct at all levels of IRE - not just 100.

Hi,

I am playing with this approach too, but to measure gamma instead. What I am doing is setting everything manually on the camera (Canon 400D or Digital Rebel XTi in the US), set a known colorspace (sRGB or Adobe RGB which have the whitepoint at D65K), and then use dcraw (command line tool) to convert RAW files from the camera to *linear* 16-bit RGB samples (no gamma adjustment, etc.) and the inspect those appropriately.

Let me know how far you get.

Hi,

I am playing with this approach too, but to measure gamma instead. What I am doing is setting everything manually on the camera (Canon 400D or Digital Rebel XTi in the US), set a known colorspace (sRGB or Adobe RGB which have the whitepoint at D65K), and then use dcraw (command line tool) to convert RAW files from the camera to *linear* 16-bit RGB samples (no gamma adjustment, etc.) and the inspect those appropriately.

Let me know how far you get.

It sounds like we are doing similar things. I will definitely share what I come up with. Let me ask you this. If I set my camera to a known colorspace (sRGB or Adobe RGB) does this mean that I can set my camera to custom white balance and I don't have to shoot a grey card to get my new calibration point for the camera? If you set a specific colorspace on the camera is that the same as saying "Camera, compare everything you are seeing against your stored profile of how white should look like in the sRBG or Adobe RGB colorspace)."

I don't know the answer for sure, but what I know at least for RAW files the whitepoint adjustment is done when converting from RAW to JPG/GIF/PNG/etc... If you stay in RAW format up to the computer, even if you set a whitepoint in the camera, it is only stored in the picture file as reference information, which you may use or not when converting from RAW to an actual image, in other words: the values of the sensors (RGBG) stay the same regardless of the whitepoint that you select.

What camera do you have? I am using a Canon Eos 400d with all settings manual, "faithful" mode, etc.

Interesting discussion. Unless you take a light reading in the room, or use a known source (eg grey card) how will you know whether or not the screen is correctly calibrated? I think having a known source in the image gives you an immediate reference point to work from.

Mark

antorsae: I am currently just using a Canon Powershot model to take a few test picture. My original idea is to have a camcorder of some sort hooked up so I get a continous feed to the computer. I am not sure if there is something corresponding to RAW when you are using camcorders...

Mark_H: My thought is that I know what color temp I want to end up with: D65. . My idea was to calibrate my camera to D65 by using its Custom Whitebalance setting. I would then take a picture of the screen. Since the camera has been calibrated to D65 it will knows how much it needs to modify the colors that it is seeing to match up with the D65 calibration. If I then display a 100 IRE screen, take a picture and then compare how "white" the pixels in that screen is I can see how close to D65 it really is. My assumption then is that if the white pixels are near 255 in value the projector has the correct color temp. If the values are not close to 255 then the projector doesn't have the correct color temp. With this idea I wouldn't be able to tell what color temp the projector is using...I am just interested to know if it is D65... I am still figuring out if this approach is viable or if my thoughts are flawed. I would then have to go up and down the entire grey scale (different IRE windows) to make sure that the color temp is consistent over the entire grey scale.

How exactly will you calibrate your camera without calibration gear that you do not have. Once you buy the calibration gear for a camera - you just as well have bought calibration gear for your PJ. Graycards only work if they are viewed under a D65 calibrated light source, not just any 6500K lamp - it must be D65. So once you buy the calibrated card and lamp - you just as well have bought calibration gear for your PJ. Next you need to buy calibration gear for your PC monitor/software/video to make sure your photoviewers are not corrupting the color from your camera. Once you buy that calibration gear for your PC - you may discover that you can get some aftermarket software to let you use it with your PJ. So then you realize you don't need the calibrated camera, the calibrated light source, the calibrated monitor/PC, the calibrated gray card. You just need a calibration sensor that works with a PJ. Calibration for photography can be just as expensive as calibration for PJ.

Get the picture?

Kras,

The big question that I have is that whether pro-sumer cameras have already color profiles to match D65K in manual mode (no whitepoint balancing). I honestly don't know the answer, but I am not too concerned right now as the primary goal for my application is to:

1) Measure brightness and color uniformity across the screen (even if the camera has a weird WB I don't care as I am comparing everything relatively to the center of the screen).

2) Measure gamma and greyscale tracking for R G B. Again, if the camera has a weird WB I don't care as WB just means mixing R G B in different proportions (across all IREs).

Mobbe, Mark... Canon has an SDK which I am using to remotely drive the camera (needs to be connected via USB). The idea is to have a computer driving the camera (shooting and taking RAW pics), analyzing them and showing results or directly interfacing with the display device for auto-calibration.

It is critical that you get RAW data from the camera, and extract it linearly, without any gamma applied to the samples. Actually digital cameras have pretty sensitive CMOS sensors so I expect to get better <20 IRE readings with a digital camera than with a (cheap) calibration sensor (Eye2One, Spyder II, DTP94, etc.).

How exactly will you calibrate your camera without calibration gear that you do not have. Once you buy the calibration gear for a camera - you just as well have bought calibration gear for your PJ. Graycards only work if they are viewed under a D65 calibrated light source, not just any 6500K lamp - it must be D65.

Does the camera need calibrating? I don't see why the following hypothetical setup wouldn't work - a projector (calibrated or otherwise) shining on a screen and a known grey reference card also visible. The picture is captured and sent to the PC. The PC examines the grey card and calculates the offset errors from grey caused by the "white" light coming from the projector and any camera miscalibration (this would be no different from using Photoshop's white balance tool on the grey card to correct the white balance for the image...). It then looks at the white screen, applies these offsets, and hey presto has the accurate values for white as currently seen on the screen...

???

Mark

Let me digress for a second: there seems to be a calibration obsession with D65K. IMHO, I think it is equally if not more important to get a nice gamma curve with SOLID blacks, good grayscale tracking (across all IREs) and not black crushing than just focus on D65K.

Another potential application of digital cameras would be to measure the light diffractions that kill ANSI contrast (I have a CRT setup and it is very obvious), model it and make a post-processing filter (in ffdshow or similar) that neutralizes it when possible (i.e. it will not improve pure ANSI contrast but it will improve detail of mixed APL scenes).

E.g. Display a NxN white box at various IREs, measure light spill across the screen and then write a program that calculates the approximated light spill for an image and subtracts it from the source image so the resulting image on the screen is more faithful to the original.

Let me digress for a second: there seems to be a calibration obsession with D65K. IMHO, I think it is equally if not more important to get a nice gamma curve with SOLID blacks, good grayscale tracking (across all IREs) and not black crushing than just focus on D65K.I agree with you (somewhat), and should note that actually, it is not 6500 color temp that is important - it is having the correct CIE x,y values throughout the grayscale range. There is a range of x,y vales that actually calculate to 6500, and you can definitely see the difference in the color of gray between them!

You are correct a flat grayscale and gamma may be more important as far grayscale is concerned because you will adapt to whatever flat tracked white you use, and can easily notice differences across IRE. But that is not why we use D65 rather than a Kelvin color temperature. By not having D65 as your white point means that CMY is offset far worse than your white point is - you will not be able to adapt to off colors like you do to white (within reason - 6500K ranges from magenta to green!). That is why we do grayscale - so that the colors within that secondary gamut are correct - because that is where most natural colors lie.

In order for you getting grayscale flat - you are relying on your cameras sensor to be linear - and I highly doubt that it is.

Agreed on the X,Y measurements.

Re: linearity of CMOS sensors, take a look at this:

http://www.clarkvision.com/imagedetail/digital.sensor.performance.summary/

and this:

http://www.astropix.com/GADC/SAMPLE3/SAMPLE3.HTM

Especially:

The recording response of a digital sensor is proportional to the number of photons that hit it. The response is linear. Unlike film, digital sensors record twice the signal when twice the number of photons hit it. Digital sensors also do not suffer from reciprocity failure like most films.

If we knew the spectral response of the R G B filters (before the light hits the CMOS mosaic array), then it could also be possible to measure X,Y from a digital camera... or am I missing something?

Then if you could get the raw sensor data since they say other than photon/quantization noise it is linear you should be able to use it for tracking. I was thinking they might be like LCOS/LCD panels which are inherently nonlinear. The problem with cheap calibration sensors is that the filters are never exact even when claimed calibrated - but if you are saying you can get raw unfiltered sensor data from a camera and do your own raw conversions - I dunno enough about cameras but does not make sense - as the filters are optical. I highly doubt that your ordinary Canon snapshot models are doing things properly or will let you get raw sensor data. To go beyond a linear grayscale and achieve a D65 you would need good filters - if not then some way to measure the actual filters inside the camera - just like when manuf try to calibrate a cheap calibration sensor - with all their lab gear they cannot get it right!

I seem to remember 'wm' who does the DILA/LCOS calibrations on a mail-in process saying he does something like that to measure panel shading/unformity- but with astronomical quality cameras and his setup does not travel and unlikely to be replicated on a budget. So maybe he can pipe in on why it would never work outside his lab.

Here's an example of the spectral response of a Nikon DSLR:

http://scien.stanford.edu/class/psych221/projects/05/joanmoh/spectral.html

As for raw data, Canon has SDKs that let you get RAW data (as well as remotely control the camera), then it is a matter of choosing the right software to apply (or not apply) the required processing to the RGGB samples.

Best regards - Andres

How do you plan to cheaply measure the filters - "We used a white-light source and the monochromator....with the Photo Research PR-650 spectroradiometer"

Your whole assumption in being able to photo a D65 relies on the camera filters being spec in the first place, and if not spec - applying some camera calibration.

For my application I was going to rely on published specs. I may have probably diverted the thread as I primarily intend to measure grayscale tracking, gamma and ANSI contrast; D65K would come later.

BR - Andres

Antorsae,

when you use the dcraw tool do you use the output color space of sRGB or XYZ? If you go to the sRGB wouldn't that transformation apply that color space's gamma to the pixel values? I wonder how the transformation is done to XYZ... If you go to XYZ it is easy to go to your [x,y] coordinates.

Reading up a bit more on color spaces I realized that my previous posting wasn't very good. I thought that outputting to the XYZ color space would give me the XYZ values that I was looking for to be able to calculate the [x,y] chromaticity values. But I am suspecting that outputting to the XYZ color space will just use the primaries and white point that is defined by the CIE 1931 RGB system. So in the end the picture will be a RGB image. I will still have to use a 3x3 matrix to go from the RBG value to the XYZ values. Am I correct in this?

RGB gamuts are defined in XYZ space - XYZ does not have any gamut restrictions other than visible color frequencies.

Your camera would have to actually use XYZ filters just like a calibration sensor if you wanted raw XYZ. But it is not a calibration sensor - it is a camera - so if you have the option to get XYZ - then likely it is processed back from the RGB capture using idealized coordinates for RGB filters - which means it is likely to have been constrained by a gamut already, and may not accurately represent what was captured. If you have the RGBW specification - you can transform from one domain to the other.

I highly recommend reading Poynton's books and articles on colorspace before you get in over your head.

GUYS!

This is not an effective idea. I know it sounds simple and intuitive to use a camera to measure a display, but in reality it very much is not possible to calibrate to D65 using this method. The reason is that the SPD of various displays differ widely, and the filter curves of various cameras also differ widely. Because of this, two colors that we may see as different, the camera may see as the same, and also colors that the camera may see as different we may see as the same, all depending on the interaction of the display's curves and the camera's curves, in addition to the response of our eyes.

In other words, you could only use a camera as an optical comparator. So you could use the camera to ensure that your greyscale was flat, but you could not use it to actually align to a specific color temperature. In order to do that, you would need to measure an identical display as yours that was already correctly calibrated and aligned to D65, and then match that measurement on your(identical) display using that same camera. You could not use any other cameras or any other displays for this(including other things like a greycard lit at D65, or a greycard in daylight or the like).

Now, another thing you could do, again with using the camera as an optical comparator (NOT as a colorimeter which you cannot do), is to have a grey card and a D65 light source, and then visually align your greyscale on your display to match the grey card lit with a D65 light source (not just a 6500k blub). Then, using your camera, you could measure the display that you have visually aligned, and then use that to make your greyscale flat across the luminance range to that same temperature.

But as for the original poster's idea, no, that idea does not and will not work in the way you intend it to. That method will not be at all inherently accurate, and you might as well just set things by eye.

Any one knows how to convert EV value of certain shutter and aperture to FTL?
There is a table comes with Minolta that you can use but I am interested in using a SLR to do lumen reading. There is a conversion fomula that I am interested in.