Serene Life

by garik

submit your photo


Hall of Fame
View past winners from this year

Please participate in Meta
and help us grow.

Take the 2-minute tour ×
Photography Stack Exchange is a question and answer site for professional, enthusiast and amateur photographers. It's 100% free, no registration required.

I'm a casual photographer with an APS-C camera (Pentax K-5) and a set of proper lenses.

Someone asked me if I could photograph his oil paintings for a catalog. It's a low priority job to him, he just likes to have something to show to others, not present the catalog in a museum.

Yet, I wonder if I can accomplish a proper reproduction of the colors if I use this fairly low-end equipment for this job, and have the images printed into a book by a professional printing service.

I understand that I need to observe the following items in particular:

  • Good lighting. Probably lots of large area lights, from both sides, ideally from all four corners. I'd probably rent those for the job. If I use multiple lights, I also have to ensure that they all have the same color, and that I dim any other light sources.

  • A color chart that I need to photograph as a sample directly in front of the lighted object, as a reference for later processing. I assume this also covers the white balance.

  • Take the shots in RAW.

I am concerned with accurate color reproduction. What must I do so that all colors get represented correctly in the final print?

In other words: I do not want to modify the taken images "artistically", I want them perfectly reproduced. I believe there is a difference and that's why the related articles on calibration do not answer my question well enough.

* Update Nov 4, 2013 *

I did not originally express my concerns with the colors well enough, so here it is:

I had occasionally read that DSLRs would have trouble with certain colors, turning reds into purple tones, for instance.

I now believe that this is caused by the internal JPEG conversion in the camera, and is not a weakness of the sensor itself. I understand that many digital cameras try to "beautify" images when developing them from RAW to JPEG, and this is probably the reason for these tone errors.

Yet, if this color issue is part of the RAW-JPEG conversion, then I wonder if the same won't happen if I use a RAW converter on my computer?

That's why don't trust RAW converters and why I wondered if a color chart is the safest solution.

All suggestions so far, however, claim that I can solely rely on the RAW converter and white balance - no need for a color chart.

Also, as there are other related articles pointed out, I like to clarify that I do not have a calibrated monitor and I do not think I should need one. I want to transfer the image unmodified from camera to printer. The only task I use the computer for is to point out the white point through the gray card, and for that I should not need a calibrated monitor. (And yes, I have calibrated my monitor, but it's a cheap monitor that can't even show the complete brightness range, so I do not trust it anyway).

* Update Dec 19, 2013 *

Here's another thing I had always somehow expected to find and now finally ran into by reading more about color temperature: The Color Rendering Index (CRI).

There are light sources that have a rather low CRI, such as LEDs, apparently (see http://lowel.com/edu/color_temperature_and_rendering_demystified.html, especially the Comparison of High & Low CRI Fluorescent Lamps).

It suggests that a low CRI will not get all colors recorded by the camera's sensor accurately. And that a simple white balance can't fix that because it can't know which individual parts of the spectrum need correction - white balance works on a much broader and simpler scale.

This means that I not only need a uniform light source but one with a high CRI. Some answers here pointed using "good" lights out (only R Hall was very particular about it, though), so it appears that this is a critical factor indeed to get the "right" lights for this. And yet, someone at Calumet recommended to use LED lights for my repro work - that's a bit confusing.

While you may argue that any light source I would probably use (including a common camera flashlight) would provide light with a high CRI, it's these theoretical complications affecting color accuracy that caused me to write this question. Even though I could not initially express where I expected problems, this is finally one example where it could affect the color accuracy, even if it's benign in the setup I'd be choosing. But I wanted to know it better than just getting a "don't worry, it'll work" answer. Maybe I should have asked "What factors can affect color accuracy" instead.

share|improve this question
1  
possible duplicate of How do I best take pictures of paintings? –  Itai Oct 14 '13 at 18:06
    
    
I think the particular subject matter here is a red herring; the question is basically about how to get accurate color reproduction with digital files sent to a professional printing service in general. –  mattdm Oct 14 '13 at 18:51
    
@Itai, as I already mentioned in my post, I read the one you point out, and I also explained why mine isn't a duplicate, i.e. because I have a question about color accuracy that the other question doesn't talk about. –  Thomas Tempelmann Oct 15 '13 at 21:40
2  
@ThomasTempelmann It doesn't really matter; the approach is basically the same. You can ignore your monitor but the coverall olor management workflow doesn't change. If you really don't care very much, though, you can assume that the sRGB and aRGB output from your camera is close enough. (That's basically what the answers so far have said.) If you do care, though, dive into the other question. –  mattdm Oct 15 '13 at 22:05

5 Answers 5

up vote 3 down vote accepted

Given your update, I would offer that color with digital photography is as much a problem of mathematics as it is getting proper illumination and white balance when actually making the photograph. Your camera senses light, separates that light into discrete collections filtered into certain ranges of wavelength (reds, greens, and blues). Depending on the exact camera, the range of wavelengths may overlap a little. The amount of overlap can affect the direct-from-camera color reproduction, however that isn't the end of things for digital photography.

R Hall offered that cameras do not see the same way humans do. I would disagree for the most part. Cameras sense light in three distinct bands of wavelength, very much like humans sense color in three distinct bands of wavelength. The primary differences between human vision and camera vision is the fact that the human eye has a fourth sensing element: rods, capable of sensing luminance with incredible accuracy at extremely high density. The human eye also senses magenta, rather than red, thanks to a double-peaked sensitivity curve for "red" cones, which changes the formula our brains use to interpret the data it receives from our eyes, but only a little bit. In general, computers can process the color from a camera in very much the same way as our brains process the color from our eyes...via a dual-axis plane: Blue/Yellow and Magenta/Green (luminance is then effectively a z-axis penetrating the center of this color plane). Discrete red, green, and blue pixel values are generally translated into Luminance, A* and B* components in what we call L*a*b* space (a color model that closely resembles the way human vision works.) Once in Lab space, we can easily adjust white/color balance, remap distinct colors, and adjust the entire matrix of "color" to produce exactly the kind of results we want. For the most part, this complexity is hidden from you the photographer by layers of advanced computer code, and presented to you as a simple interface...maybe a slider for color temperature and a slider for color tint, or a series of RGB curves, or even simpler...a camera profile that you can simply select to apply the right set of curves and other settings to correct for distortion and the like.

Once you have a RAW image on a computer with RAW processing capabilities, color reproduction is really up to you. Color is actually reproduced via mathematic algorithms applying RGB tone curves and white point adjustments and tonality shifts to the interpolated RAW sensor data. You can tweak those tone curves to your liking, either directly if you have the software, or indirectly by using color profiling. With some basic color profiling, using an X-Rite ColorChecker chart and a known illumination with a known white point that will be reused to illuminate the paintings you later photograph, you can create a custom color profile to accurately reproduce the colors of your paintings.

A RAW converter is simply a starting point. They don't ultimately dictate what happens to the red, green, and blue pixel values in your .CR2 or .NEF files...you do. You can manually tweak color with the RGB color curves, or you can create custom color, camera, and lens profiles to extract the maximum amount of color and detail accuracy you want from your photography. Once you calibrate your software, it should be easy enough to simply import, resize, and print, without actually calibrating your monitor.

Personally, I would highly recommend you calibrate your monitor because that is really the first place you actually SEE your work, and also the first place you will be able to identify any major color discrepancies vs. the original paintings themselves. You could certainly skip that step and simply print...but you could burn through a fair amount of print materials (which are far from free) before you actually get your color correction completely worked out. As such, it is fairly important to maintain an accurate, corrected workflow for managing color correct image processing...calibrating your monitor should save you some money in the long run. (To note, when I have properly calibrated my monitor, factoring in the local ambient lighting, I can lift a print up to my screen under that light and the results are VERY similar. A computer screen has a deeper black and a brighter white, but outside of that color reproduction accuracy should generally be obvious.)


I'll put this here temporarily, but it would be best added to a question that explicitly asks about color profiling, which depends on proper illumination. If you are performing photographic work that depends on maintaining a color accurate workflow, the first thing you are going to do is illuminate the scene with the correct kind lighting. Once your scene is lit properly, you will then need generate a profile, for that illuminant, and save it on your workstation for use while processing future photos created under that same exact illuminant.

So first, illuminating your scene. You are correct, the average CFL and even more so, the average LED, do not produce a quality spectral power distribution to fully and accurately light your scene. CFL bulbs are better than LED these days, however they still tend to concentrate color in one band or another, without offering the broad spectral distribution that will ensure that all wavelengths of light illuminate your scene. Why is it important that you illuminate your scene with all visible wavelengths of light? This image demonstrates the SPD of various light sources, including daylight:

enter image description here

If you compare the SPD of a low pressure sodium lamp (the kind of lamps normally used to light our highways) with that of daylight, you can see the problem. Low pressure sodium is a narrow band emission, only emitting high intensity orange light. It lacks the majority of the rest of the visible spectrum. Mercury lamps are not much better, although they do emit light in spikes across a broader spectrum.

The problem with "spiky" SPD is that you get a lot of certain wavelenths, and littler or no of most wavelengths. Since photography is based on reflected light, in order to accurately capture all of the color and detail in an object (such as a painting) it is important to ensure that the illuminant used to light your scene is broad spectrum, one that offers a SPD that is less spiky and more uniform, across the entire range of visible wavelengths. No artificial bulb will offer the broad intensity of color as daylight, however a good High CRI bulb will produce a more balanced SPD with more intensity across the full spectrum, and usually a couple spikes, around yellow-orange and blue. For color correct workflows, a CRI of 98 or higher is ideal, and preferably one of fairly high wattage to ensure you can use a low ISO and high shutter speed.

Once you have a proper broad spectrum illuminant, you will need to perform some color calibration. Color calibration is actually pretty easy these days when using something like a ColorChecker card and companion software (you can get these from X-Rite). All that is really required is to place a standards compliant ColorChecker card under your illuminant and photograph it. Once photographed, you import the images of your ColorChecker into the companion calibration software and generate a profile. Such a profile can be used in a variety of software (such as Adobe Lightroom) to perform color accurate RAW import and conversion.

enter image description here

When creating a color profile with a ColorChecker card, it is best to have the same card out and visible, even hold it up next to the screen with the photographed copy visible on screen (make sure your workstation area is illuminated by the same High CRI light). You can visually compare and contrast the colors of the card with the colors on screen. Any significant discrepancies will usually jump out at you. If you do see any discrepancies, you either have the option of manually tuning the tone curves for the calibration, or trying again with a separate set of photos. In order to perform color checking in this manner, you will need a properly calibrated screen. You don't necessarily need a high end professional grade screen to do this, however you will need at least an 8-bit screen (rather than a 5- or 6-bit screen, which tend to be th cheapest, often used for gaming for their high response rates.) Ideally, the screen would be IPS type.

Once you have used a ColorChecker card to profile your workflow, the rest should be largely "automatic." When you import your RAW images, apply the custom profile. If you need to do any basic tone and exposure tweaking (i.e. recovering highlights), do so. If you use a tool like Lightroom, once you make your basic edits, you can save them as a user preset, and simply apply that preset on import to the entire bulk of your photos of each and every painting you have to photograph. Once imported, you can pick and reject, then bulk export your picks to TIFF for further processing....or simply print each of your photos directly from Lightroom. After color profiling, your workflow should be reduced to a very simple "import, pick, print" procedure (which, as I gather, is what you are looking for.)

share|improve this answer
    
Updated my answer with the context for my comparisons between human vision and cameras. –  R Hall Dec 19 '13 at 3:09
    
I have calibrated my monitor (which is a cheap one that doesn't even cover the entire brightness range!), but I still claim that this is useless in repro work because I do not even remember the colors when I post-process the images on it, so I cannot "adjust" them when I cannot tell what the colors were in the original. And at the place where I'd do the repro work I only have a MacBook Pro with me, and its screen is even worse that the Dell monitor I use at home, so even comparing a painting right there is probably not very accurate. –  Thomas Tempelmann Dec 19 '13 at 9:09
    
BTW, I appreciate your effort writing this long reply. However, it's too general for me. I have added a new "CRI" section to my question. If you could comment on that, it would great. I realize that I keep dragging this along but as an engineer, I like precise solutions, and so far it's been all rather "oh, just calibrate your monitor, trust the RAW converter, and nothing can go wrong" like answers. I like to understand it, not just follow advice blindly. –  Thomas Tempelmann Dec 19 '13 at 10:46
    
@ThomasTempelmann: You might be better served by more explicit questions, then. I could edit this, but I think you could get more explicit and precise answers of you asked explicitly about just CRI for illuminants, and just how you can ensure color-accurate profiling of your scene, etc. You can keep expanding your question indefinitely, but then this whole entire topic will ultimately lose it's relevance. –  jrista Dec 19 '13 at 15:15
    
A Color checker may not provide accurate color management for paintings. The 24 colors were chosen to represent scene colors and pigments, and do an "ok" job on those. But have proven to produce inaccurate results for other types of pigments such as those in paintings. Roy Burns has been working on this issue for many years . Here's a link to some older foundational information that may provide a better perspective: art-si.org/PDFs/TechnicalReportNGAPrint.pdf –  R Hall Dec 20 '13 at 5:03

From my experience, accurate colour reproduction comes from getting it right in camera first.

This is how I would approach the scenario:

  1. Set the camera up properly - Use a lense that minimises vignetting and other unwanted distortions. Make sure it's of a focal length that means you're not of a distance that is uncomfortably close to the canvas. Shooting in RAW will give you maximum flexibility when it comes to post-processing, particularly with the White Balance (which will have a knock on effect for colour reproduction).
  2. Use a tripod - In conjunction with a remote shutter release, this should minimise camera shake, resulting in a sharper picture.
  3. Proper lighting - How you approach this is up to you, but even lighting will help improve the picture.
  4. Get the exposure right in camera - The closer you can have to "perfect" exposure, the less work you have to put in to post-processing. It is my experience that the closer the exposure is to being right, the more accurate the colour reproduction will be.
  5. Post-processing - Like it or not, shooting in RAW requires some form of processing. As you want your images to be colour accurate, you will need to calibrate your monitor so that you can be confident that the final image you give to the printer will be accurate. You will want/need to ask your printer what format he would like the images in (JPG/PNG - I'd suggest PNG as it is a lossless format), what colour space (sRGB/AdobeRGB - my understanding is that sRGB will give you more consistency regardless of whether you distribute it through the internet or through print media), aspect ratio (this depends on the format you're printing to - 10x8/4x6/5x7 etc). Essentially, I don't think that a printing company will accept a RAW image straight from your camera, and by giving a completed image you're saying "this is what I want you to print, and it is true likeness to the original".

It appears that the normal way of accurate color reproduction relies on the RAW converter to map the colors into a standard space that is understood by the output medium (here: printer). But is that really the best and only technique for reproduction? I'm just saying: I don't trust those RAW converters to do the best job at this. Or should I? (For instance, what about variations in the sensors, e.g. by temperate shift or aging, couldn't they affect the color recording? Also, if my lights are not perfectly white, won't that cause more than just a white balance shift, which the RAW conversion software can't take into account?

Why don't you trust RAW converters? Tools such as Lightroom are used by professionals the world over and when used in conjunction with software such as Photoshop, result in images you see every day online, in magazines and in advertising. These programs have sophisticated algorithms which will allow you to correct for any faults in your camera or lense. You can correct the white balance, the individual colour levels, the exposure...

Ultimately, you're not going to want to modify the sensor on your camera, so you will need to ensure you manage the variables to get as close as you can to the original. Whether you do this in preparation for taking the image or post-processing is up to you; although I'd recommend both to ensure the best quality pictures.

For what it's worth, I have a Pentax K-5 II and have no qualms about its ability to capture colours accurately.

On a side note, don't brush off the other answers. By having calibrated monitors and an understanding of White Balance and exposure, you can improve your photography skills and know when to artistically under/over expose your images.

share|improve this answer
    
Welcome to Photography on StackExchange. Great answer. I'd also add that the RAW tools, particularly if you use the ones provided by the camera manufacturer, do the same things as what the camera does, it just allows you to do a better job than what the camera will do at guessing. Ultimately the image will always have image processing applied, working with RAW just lets you control what image processing is applied. –  AJ Henderson Oct 17 '13 at 13:43
    
Thanks, but I think most of the answers are boilerplate, i.e. they've been answered already in the linked posts (plus, the important part you should mention about lighting is that it needs to be all of the same color!). Though, repeating them can't hurt, either. However, I do not see why a calibrated monitor is necessary if the picture as taken by the camera and then processed by a RAW converter and then brought to the printer needs any (possibly poor) represenation on the monitor. After all, aren't the colors calibrated by the RAW conversion already, and those are used by the printer? –  Thomas Tempelmann Oct 18 '13 at 10:22
    
@ThomasTempelmann When you shoot in RAW, you have to convert that file into a JPG/PNG. Part of that process is setting the White Balance. If your monitor has a potentially poor representation, how do you know that you or the computer are setting the WB properly? –  Aiden G Oct 22 '13 at 17:52
    
Aiden, that's my point - I'd use a color or grey card as the reference, and then I tell the computer that to use that for the white balance - even if my bad monitor is not calibrated, it's still correctly adjusted with the reference card. That's the hole point of this question. I do not trust my tools (computer, camera sensor), that's why I wanted to know what I can trust instead to get the colors right. –  Thomas Tempelmann Oct 24 '13 at 9:26
    
@AidenG, see my update about CRI. That's what I had in mind: That there are cases (though, unlikely when I use "good" light sources) where a simply white balance does NOT fix the colors. Also, the RAW converter would not know that it would have to correct certain spectral areas individually to correct such a bad light source. That's what I meant with not "trusting" it. I meant that I may need more control over what the RAW presets can do. Hence my suggestion of a color chart that would probably deal with such low-CRI errors. But it's theoretical, not really practical issue, I admit. –  Thomas Tempelmann Dec 19 '13 at 10:57

Here's the problem with photographing paintings: Cameras are RGB devices that don't see the way humans see. So there is going to be problems matching colors in a painting, because a camera that's profiled using a target like a color checker, is most accurate on those pigments and colors...in the color checker.

There is a scientific phenomenon called the Luther-Ives conditions that I am referring to when I say cameras and humans don't see the same. What this well known condition describes is the conditions that RGB devices must meet in order to see like humans. When the Luther-Ives condition is satisfied, a camera sees color the same as humans. Meaning the spectral sensitivities of the RGB sensor are the same as human vision. When that happens, (it never happens) then cameras won't require as much color processing to correct. As an example of this, without special filters and coatings todays cameras will see infrared light. Thus the reason for ICC profiles and color management.

So now comes the fun part. If you must get good reproductions, then you need to make a target from the pigments used in the painting. If you don't, it's very possible to have some colors reproduce wrong or even the same as other colors in the painting,

The other part of that puzzle is lighting. Light is color, so lighting with a constant color temp and good quality studio light is essential. So at least Paul C, Buff's Einstein's or high end Pro light with adjustable color temp.

Paint your target in the same format as a ColorChecker and if possible given the color pallet you shooting, the same position and colors mixed from the paints in your paintings. Measure that using a spectro to get reference values for your target. a ColorMunki will work for this if your going cheap, otherwise an i1Pro2)

Set your lights to 5000K and shoot your created target. Use that image as input to your camera profiling software and choose the reference file you created with the spectro. Now you have a profile for your camera that will get you as close as possible without spending a bunch more time and money.

If the above is overkill for your process and your looking for just OK, use a color checker passport system and you can get from pretty good to not bad results, depending on the painting and pigments.

share|improve this answer
    
I can rent an i1 color calibration device from Calumet (where I would also get the lights). I've used the ColorMunki before, but its software only works with monitor screens and printers, both of which it can control. How would I use that on a painting that already has its own colors? I guess I'd need different software than that what comes with the device. I guess I would then give the software my image from the camera and let it see certain colors on the painting so that it could match them. Not very practical with already finished paintings, though. –  Thomas Tempelmann Dec 19 '13 at 9:03
    
As I mentioned, make a Color Checker with the pigments used in the painting. There are many ways to discover those pigments, If you don't know them. Asking the artist usually works...unless their dead. ;) –  R Hall Dec 19 '13 at 12:52
    
@ThomasTempelmann: Your comment above probably also warrants it's own question. Keep in mind, calibrating a device is intended to calibrate THAT device, and that device alone. You don't need to think about calibrating all the devices together, just calibrate each device independently. Once calibrated, ICM deals with maintaining accurate color throughout your workflow. When it comes to print, you actually calibrate a the printer+ink+paper (and, if you want to get really technical, +viewing light.) You can usually find ready-made color profiles for paper brands for major printers (i.e. Epson). –  jrista Dec 19 '13 at 23:33
    
@RHall, that issue with matching the pigments correctly sounds reasonable. Isn't that also the "illuminant metameric failure" effect I mentioned earlier? Also, how do you suggest I make a color checker from existing paintings? I cannot pinpoint the colors in a painting exactly for measusing - it's simply not practical. And I cannot paint a new picture because the paint color are not available any more. –  Thomas Tempelmann Dec 20 '13 at 14:16
    
@ThomasTempelmann What that is is creating metameric match. That's what good color management does. Illuminant metameric failure happens after a match has been made under one illuminant. So If your not able to measure the painting and get the spectral reflectances, and determine from them what pigments were used, or research them with the artist, then your down to the degree of success your best process can produce. You may have to color correct for hours to get the final image to look enough like the painting to sell the process. It all goes down hill from the best case scenario. –  R Hall Dec 20 '13 at 15:37

Do not worry about color accuracy. The colors might probably end up to be slightly off, but the problem will be almost invisible.

The biggest problem is the white balance, as seen in the referenced posts. Once you treat this correctly, you are on the safe side. This means that the light color must be uniform, i.e. do not mix almost sufficient fluorescent light with additional flash.

Color chart is very good point, but for slightly different reason. Get one, which is widely used and put it under/besides the painting and make photo, so you'll have both of them on the image and then crop the needed part. The raw conversion software is expected to do very decent job. Additionally, you will probably produce image in some RGB colorspace variant, whereas for print it will be converted to CMYK (but this is not your business).

But having the photo with color chart, the guys at printing company do have the correct reference, so they can set up everything properly. They have all the professional color meters and stuff, so even if your RGB image won't be perfect, it's fine to state: "I want the colors to match this color chart" and they should be able to achieve that as a part of the print process.

share|improve this answer
    
-1 This is not a reasonable answer. When the OP asks "How do I ensure X?", the correct answer is not "Do not worry about X." There are certainly questions when someone may be overly concerned about something that really doesn't matter, but accurate color reproduction in a photography/printing context is not one of them. –  Caleb Oct 16 '13 at 15:31
    
I, however, think that this answer did indeed help. He confirms how using a color chart migth be helpful, and that I do not have to worry too much about what I thought I have to worry about. He essentially understood that I'm unsure and that I needed advice. The fact that this entire question of mine is maybe too vague is another issue. If you think, however, that his advice is off, then please add your own answer, as I do not have the experience to decide how good his advice really is. –  Thomas Tempelmann Oct 17 '13 at 8:59

I've talked to someone who writes software that deals with color management and understood my worries exactly.

In summary, he suggests:

The most important part is using a high quality gray card. The cheaper ones may suffer from illuminant metameric failure, i.e. they are not ideally neutral to the light that they reflect.

If I plan to make single (expensive) prints, a color chart may be used as well, which I would then also give to the printer, allowing him to calibrate his machinery. Though, for my above described purpose, this is overkill, and the RAW development should be sufficient as long as I keep the white point (identified using the gray card) properly adjusted.

share|improve this answer
    
Illuminant metameric failure is not a factor in adjusting neutral balance in a camera, or matching paintings. The definition mentions the cause is lighting changes that cause two different samples to match under one light to be different under another light. The reason for that is the make up of the cards. If they both had exactly the same reflected spectrum they will match under any light. Problem is your not shooting a card but a painting. Lots of pigments in a painting, none that match gray cards. –  R Hall Dec 19 '13 at 3:16
    
@RHall, okay, but what about CRI, see my updated question. I am still not convinced that a simple white balance can fix all kinds of color inaccuracies, and what I just learned about CRI supports that. Of course, as you pointed out, a good light source would not cause this problem. I just like to get some confirmation on my findings. –  Thomas Tempelmann Dec 19 '13 at 10:48
    
Good controlled white balance is just a factor, not "the factor" as I mentioned light is color. So that must be controlled before profiling. Many times I visit a studio and notice their lights are not even set to the same color temperature. So I did not say good white balance is your solution, It is not. –  R Hall Dec 19 '13 at 12:52

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.