cgardner
Offline
• • • • •
Upload & Sell: Off
|
Warning: Long and rambling essay on gray cards and digital capture vs film.
The guy who makes the WhiBal cards did a good bit of testing and claims that a lighter shade of gray than 18% produces a more neutral WB in the technical sense of reproducing the card correctly, but you need to keep in mind that setting neutral WB is a process control aid rather than a desired perceptual result for pretty pictures of real stuff.
If shooting RAW setting Custom WB does not affect how the image is captured. What you control by setting WB in the camera by any means when shooting RAW is how the images from the camera first hit your eyeballs PERCEPTUALLY on the monitor. We set custom WB they have a consistently neutral baseline WHEN FIRST SEEN.
First impressions are important in color because the eye will adapt to the monitor just as it does to the scene in real life. It is possible to just shoot one frame with a gray card, click correct it in ACR, then batch paste the WB change to the rest of the files and get the same net result, but it requires more steps. If custom WB is set on camera the files are balanced and ready to view from a neutral baseline immediately.
Which method of WB works best is an workflow issue. Sometimes its impossible to do either a custom WB or put a card in a test shot. The next best approach is to just pick the closest pre-set and stick with it. The worst possible approach for process management is AWB - auto white balance, which resets WB for each shot based on the assumption the brightest areas are neutral, which isn't always the case.
Having a known baseline is a fundamental principle underlying any form of process control: to control any process you need to first define what is "normal" to identify the exceptions to the norm. In most processes there is a range of deviation from the norm which will still be considered acceptable performance. Color perception is no different. Its not absolute, but more of a moving target affected by context and other factors.
In the digital realm the process control assumption everything hinges on is that a neutral gray object has RGB values which are equal. That's why its important to start the process control process at the camera capture stage with a test target known to reflect RGB equally over a wide range of lighting conditions (i.e. metamerically accurate). If you have a file with equal RGB values but it doesn't look neutral on your screen, because the file is correct you can then deduce your screen must not be calibrated correctly. Once you get your monitor so it is displaying color neutrally, you would expect a photo of a neutral gray card captured with your camera to appear neutral on the screen. If it doesn't then you can deduce the camera did not record it with neutral WB due to the way WB was set on the camera to match the ambient light.
The gray card reference is needed because human color vision is very adaptive. That's why you don't see the green bias in the light under trees, and will not see it in the photos you capture under the trees either on the screen if you stare at it long enough for your eyes to adapt. But if you also have a known neutral gray card in the photo you could measure the RGB values and correct them to make them neutral. By correcting the card all the other colors are corrected the same way and the skin tone under the trees will change from the dull grayish green captured accurately by the camera (relative to the light of clear sky at noon) to the healthy pink we expect caucasian skin to be perceptually. Setting custom WB off that same green light under the trees will provide an editing program like Photoshop with the information it needs to correct the color back to neutral automatically, without the need to click on a gray reference in each photo. It allows one to shot a reference frame of the card and have all the photos in the same light look neutral when first opened for evaluation.
Why don't we notice the skin is green in the photo taken without custom WB? Since skin comes in many different shades we don't get hung up perceptually on what exact color the skin is or the tree behind the face for that matter. We assume a caucasian face falls in a range of pink tones and accept it as rendered as falling in that range. The green bias alters the color relationships and will make the lighting on the face seem dull, gray and flat. Green + magenta = gray. At the same time the green bias also makes the foliage look more saturated is more appealing. I wish I had a dollar for every outdoor photo I've seen which had a green cast the photographer and a dozen others giving C&C just didn't see because they didn't know that gray dull skin was the clue that the problem was a green cast in the WB. Add a bit of magenta to the overall color balance and the fog is lifted and the contrast looks normal along with the color.
Perceptionally, unless we know the person and have a memory of what color their skin actually is or compare the actual photo reproduction to the skin most will accept a rather wide range of skin color as being normal. Even in person the skin will look different, depending on the light, the color of the clothes they are wearing and the background surrounding them. Color is a moving perceptual target and aiming to hit some theoretical bulls-eye often can miss the more important holistic objective: making a photo seem real.
Making a photo seem real is mostly about getting memory reference colors correct. If you were to look at a photo of a car in a photo on a white background and the paint had a slight reddish cast your eyes would decide whether or not the car is actually red or not by first keying off the white background and then off things you assume to be neutral on the car, such as the black tires or the reflections off the chrome. If the background is neutral white your initial assumption would be the car is red, until you happened to notice the tires look reddish. Then your brain will figure out the car is actually neutral but the WB of the photo is off. Conversely if the car was perfectly gray, but the WB of the background lighting was slightly warm and reddish, your eyes adapting to the expectation that the background is white will perceive the car as having the opposite bluish hue.
Perceptually color is moving target. Technically neutral color set per equal R=G=B values on a gray card is seldom what looks best perceptually for real life scenes. 9 out of 10 humans would think a portrait with slightly warmer WB would look more natural. In some situations where the ambience of the original scene triggers the emotional reaction (e.g. the "golden hours") you wouldn't want to set custom WB at all, instead setting the camera to "daylight" WB so the camera records the ambience relative to high-noon, which is what your brain is using as its perceptual frame of reference.
The main function of using a gray card for custom WB is so all photos have the same baseline for WB when opened and seen for the first time on screen. So even if a gray card used as a reference isn't perfectly neutral in the colorimetric sense, it can still be used as standard frame of reference providing its color isn't too far away from neutral. So given the choice of paying $160 for a target that gets the color neutral with 99.9999% accuracy or $20 for one that does it with 98% accuracy which is the better choice? The extra 2% in colorimetric accuracy in the card matters little in practical terms because your eye upon seeing the photos set with either using custom WB will adapt immediately and not see the small difference. It would only be noticed if there was an image of the actual card which you were able to click correct. Would one be better than the other perceptually? That would depend a lot on what was in the photo.
Many ask "If its white balance why do you want to use a gray card?" The reason you want to use some shade of gray rather than white is because with white it is possible to overexpose just one or two channels of the three which would skew the WB. Gray puts the card down in the range were an automatic exposure system will render it down in the middle of the dynamic range of the camera. But if you simply change to M mode on the camera you can reproduce the card as any tone from pure white to pure black by controlling the exposure hitting the sensor. I suggest doing just that in one of my tutorials as a way to understand how a histogram works: LINK
http://super.nova.org/TP/HistogramTest.jpg
The "tonal map" of the histogram response for my 20D above was created by bracketing exposure of a gray card (after setting custom WB off it) to reproduce it from "paper" white to as dark as my camera f/stop range allowed. That helps me interpret what tone the right edge of the histogram represents and also told me what my actual real-world camera DR is: about 6-1/2 stops of detail I can see.
So if you think a 12% or 15% reflectance will produce better WB just expose the 18% card a bit more manually or with + EC and it will be reproduced lighter! As long as the card is neutral the light actually hitting the sensor will be the same, no?
I use a "belt and suspenders" approach: Set custom WB off the card, then shoot the card in the scene when possible. The first step sets the WB, the second allows me to verify perceptually and with the eyedroppers in Photoshop that it was set correctly. The ability to actually verify WB off a target in the capture is where using a gray card differs from the "magic coaster" ExpoDisk approach. The ExpoDisk, pointed at the light source from subject position (i.e. like an incident meter) can do a very good job of setting neutral WB, but unless one also has a gray card to put in the scene there is no way to verify WB was in fact set correctly.
I'm a process control oriented, baseline kind of guy so what I do is first put my camera into daylight mode outside, or tungsten or fluorescent inside (depending on dominant source) BEFORE shooting the gray card to set WB. That isn't necessary to set custom WB, but gives me a visual reference when editing how much different the actual light was from the pre-set. Then immediately after setting custom WB I shoot the card again. That shot when checked in Photoshop should have equal RGB values and not change color if clicked with the "snap to neutral" eyedropper tool. The comparison of the two frames helps me to understand, in retrospect, what the color ambience of the actual light was relative pre-sets and that helps inform my editing decisions on whether to leave the color technically neutral or take it warmer or cooler perceptually while editing.
As for the origins of the Kodak 18% card and its historical use? That is something which I've researched.
The gray card exposure technique was developed back in the days of film negatives and simple averaging meters. A photographer would do a test, metering off the gray card with the averaging meter, then adjust the ISO setting of the meter until the indicated shutter/aperture readings produced shadow detail on the negative. Once the film ISO on the meter was adjusted that way (i.e. calibrated to the shadow detail) the all the photographer needed to do to get the film exposed correctly was to hold out the card with one hand and meter it with the other. It is important to understand that getting the highlights correct on a B&W print wasn't controlled by exposure at all, but rather by how long the negative was developed. Longer development meant more negative density which made tones on the print lighter.
A guy named Ansel Adams, building on the work of others at the time, developed a very systematic way to predict what negative development time would be needed for any range of contrast in the original scene to fit it onto the fixed range of a #2 print paper. The entire Adams zone system process was based, at the time it was developed, on that average meter reading off the card.
As far as I can determine from anecdotal accounts the 18% card actually had its origin in graphic arts where the white and gray sides were used as exposure targets for halftone dot placement in halftones and color separations. Litho film is either clear or solid black and a gray scale target becomes a form of exposure meter. As exposure increases more and more lighter tones get reproduced as white. The reason the 18% value was chosen was lost to the ages, apparently even within the corporate memory of Kodak which made them, but some it seems that there was some numerical relationship between the number of stops difference between the white and gray sides, and from my experience shooting line shots and halftones on graphic arts cameras that grey value is the target value used when shooting a line negative to ensure proper litho negative density in the white areas of a page and for placement of the 50% dot when making a halftone. So based on my experience in graphic arts anecdotal references about its origin in photo lithography makes sense because back in the 1920s when it is said to originate there were no other controls available other than a test target and your eyeballs. That was still largely the case in the early 1970s when I did conventional halftones and separations.
Regarding the use of the Kodak card for color evaluation it wasn't used for that purpose in the early days of color separation because unlike RGB color you don't get neutral gray when C=Y=M. Instead old timers balanced color with a gray card by knowing the relative percentage dots of C, Y, and M would make the card look the same with the paper and ink being used. Since direct eyeball comparison was being used it really wasn't critical that the card be absolutely neutral, only that the color separation reproduced it accurately. Its only when color management entered the realm of doing all the manipulation numerically with complicated math that its important to start with a reference which does reflect equal RGB wavelengths. Electronic color management starts with the assumption that all gray values are neutral (i.e. the L axis of Lab space).
In fact it wasn't until recent years that Kodak made its cards neutral, causing the makers of some other white balance devices costing 5x more to claim theirs were better. The "Kodak" cards are actually now made under license by Tiffen, the filter maker and have been found to be neutral. While cheaper their down side is that they are prone to damage and color shift as they age, so as a long term investment one of the plastic ones are a better choice. Search B&H for Gray Card and you'll find a wide price range. My approach when faced with the choice of $20 vs $60 piece of plastic with a raw material cost of 50 cents is to first try the cheaper one first, in part because I just view the custom WB off the card as a starting point.
You'll notice when shopping for gray cards you'll see "warm" and "cool" cards. Electronic WB had its roots in video before still photography and it was a common practice to use a WB target with a blue bias to create a warmer tone in the video. When a cooler look was desired a card with a yellow bias would be used. Setting the WB off the standard target ensured the color would be consistent for all the different takes of the same scene: wide, medium, close-up.
Ansel Adams popularized the use of the Kodak 18% gray card in still photography, but for exposure. I learned photography using his system and have a very good understanding of how it works and how the perceptual and technical aspects relate to today's digital realm.
Adams picked the Kodak 18% card because it was the only standard reference commonly available. It was already in wide use as described above in the first paragraph, and Adams liked it because perceptually the 18% tone seems to falls between white and black. Most think zones = f/stops but they are actually an arbitrary division of tones on a print, not f/stops in the scene. Read Adams "The Negative" to understand what the zones represent and why he picked 18% gray to anchor the system.
Legend has it that when Kodak realized that 12 - 13% would be a more accurate middle tone for its gray card and considered changing it Adams threw a tantrum, went to Rochester and bullied Kodak into keeping its card at 18% so as not to upset his zone system which was based on Zone V = 18% gray reflectance = middle of the perceived tonal scale. If you buy Kodak card set (Publication R-27) and read the instructions you'll find they recommend adding a correction factor to a gray card reading to obtain correct exposure.
You can use a gray card to calibrate a hand held meter for digital. For example with a Sekonic L-558 spot meter if you read a card in flat light and set camera per the meter the card will be reproduced as some shade of middle grey and the spike on the histogram may or may not fall directly in the middle. If you place a white towel next to the card may or may not be exposed correctly. But by adjusting the meter compensation factor you can adjust the meter so when you meter off the card either the spike falls dead center or alternately so the the highlights in the towel are exposed optimally as in this baseline test shot:
http://super.nova.org/TP/DR_FlatLight.jpg
But metering off a gray card to precisely control highlight exposure makes about as much sense to me as scratching my belly if my nose itches. What makes more sense is to just meter the towel with the spot meter then find the amount of exposure compensation needed to reproduce it correctly. In other word use the EC adjustment function of the hand held meter to move the default reference point for exposure from middle gray to textured white. Doing that would require EC of of 2-1/3 to 3 stops (depending on sensor DR) but once compensated that way the L-558 owner could simply point the meter at a textured highlight in the scene and read the shutter/aperture needed correct exposure off the readout without all the mental gymnastics of interpolating off the middle tone.
The same approach could be used with a camera spot metering. Once you know how much EC is needed to make a spot metered textured highlight reproduce correctly you could dial in the necessary amount of EC, point the spot at the brightest textured highlight and get correct exposure. The advantages of a hand held meter is that it is more precise, the disadvantage is that it can't take in account factors like lens flare.
Getting good results with digital is simple: just don't blow the highlights. Shadow detail? If you get the highlights exposed correctly the shadow fall where the limited range of the sensor puts them and the only way capture more shadow detail is to: 1) add more light to the shadows (i.e., reflector or fill flash), or; 2) make bracketed exposures on a tripod and blend them with masks in Photoshop.
Histograms provide valuable feedback on when over / under exposure is occurring but not where. I find its much simpler skip all the fancy metering and obsessing over what to measure, how to measure it, and how to interpret the results and simply use the most valuable exposure guide ever invented, the OVER EXPOSURE WARNING in the playback. It really makes exposure a no brainer because you need to be blind not to see when and where the over-exposure is causing clipping. When you get the exposure correct in the highlights using feedback from the OEW, then the left side of histogram will reveal if shadow detail is being lost (if the graph is running off the left side. Correct exposure looks like this on the ends of the histogram:
http://super.nova.org/TP/TowelGary.jpg
What the histogram is doing in the middle really isn't critical at the capture stage because the middle tones are easily moved around with the middle slider of Levels or similar controls to change the internal contrast to compensate perceptually for how a camera records them differently than we think we saw them in person. For example, below is an ambient only backlit scene exposed for detail in the highlights of the white towel. But due to the limited range of the sensor the middle tone and shadows are darker than see by eye: it is technically perfectly exposed in the RAW file but look perceptually incorrect and underexposed. A big reason for that perceptual vs. being there in person is because in person when your focus shifts in person from foreground to background your eyes will adapt to reveal more shadow detail. That will not happen to the same extent when looking a the dark area of a photo unless brighter areas are masked off: why the convention of putting white border on a dark photo are really not a good idea perceptually.
http://super.nova.org/TP/DR_Backlight.jpg
So I tuned on high speed flash and let the camera do its magic:
http://super.nova.org/TP/DR_FlashFill.jpg
The highlight are still the same, but the fill in the foreground renders the scene more like my eyes perceived it. The background is still dark because the fill didn't reach there, but as in person, when focused on the chart in the foreground the background is tuned out perceptually and the fact it has no detail isn't perceived as a defect. In terms of making the chart the center of interest the darker background makes it contrast more and is actually a good thing ascetically. The foreground may seem overexposed perceptually, and I wouldn't disagree. That is just on turning on the flash set to FEC =0 rendered it in the test series. But when just the card is selected and the histogram examined it shows the foreground is, technically speaking, correctly exposed or even a bit underexposed:
http://super.nova.org/TP/DR_FillHisto.jpg
The point of showing that exercise here is to demonstrate that the optimal technical parameters (i.e., perfect histogram or color balance) often will not produce and accurate perceptual rendering of the scene as by eye. There is very little which can be done to affect the appearance of the captured scene when it is being captured as with the zone system, so digital is mostly "Expose for the highlights - add flash to add shadow detail as necessary, then fix everything else in Photoshop".
The rude awakening negative film / print shooters got when switching to digital was the last part; finding they needed to learn a new skill set akin to learning how to make a photo print in the darkroom: icc color management and process control in Photoshop. Most don't realize the photo lab has been fixing their WB and over-exposed files: a color neg. can tolerate nearly 2 stops of overexposure before the color layers get out of whack.
Those who previously shot transparencies before digital were far more disciplined and conditioned to the digital workflow. Transparencies require the same precise highlight exposure, and it was necessary to put filters on the lens when shooting to correct the color temp. A pro shooter's color correction filter kit was analogous to a gourmet chef's collection of knives. If anything a person accustomed to shooting chromes with cc filters found digital far easier.
There is now a generation of camera users who never shot film, who from my perspective try to make the process seem far more complicated in nearly every respect from exposure to color than it really needs to be perceptually. I'm not anti- control. Just the opposite in fact. But I'm pragmatic and experienced enough to know the point where technical obsession doesn't yield any practical tangible results on a perceptual level.
I've found by testing that when the highlights in a textured white object start to clip the file is 1/3 stop overexposed. So if shooting a portrait I'll have the subject hold a white terry rag -- the rag rather than a card allows me to evaluate texture - raise exposure until the rag starts to clip in the OEW, then back off the flash power or aperture (if shooting manually) or FEC (if shooting ETTL) by 1/3 stop. Using the OEW and a white reference point really makes correct exposure a no-brainer even on white backgrounds:
http://super.nova.org/TP/WhiteBGTowelCard2.jpg
So for a very simple workflow, set custom WB off gray card when possible and appropriate to scene context to create an automatic neutral baseline for comparison, and then include that same card in test shots when possible. If you can't set custom WB, then at least try to capture a shot of the card or some other object in the scene you perceive is neutral. If you can't do either pick daylight, tungsten, or fluorescent; not AWB which changes shot-to-shot and offers no consistent baseline.
You may not actually want the photo to wind up technically neutral but it will give you a consistent baseline for evaluating the color by eye. Unless you have the tools to precisely measure color numerically don't obsess over numerical nuances of WB, instead trust your eyes. Try different workflows and pick the ones which offer the best results with the most convenience. It's all a matter of convenience and cost vs quality, otherwise we'd all be shooting with 8 x 10 view cameras.
Understand the difference between when a file is correctly exposed on the highlights and how at the same time the range of the sensor can make it appear underexposed perceptually in high contrast lighting. For exposure just rely on the over-exposure warning and don't clip anything except specular highlights and light sources such as street lamps in the photo. Buy a TTL flash and a bracket and learn how to use flash to seamlessly correct the fact your sensor is incapable of recording with a single exposure.
In the post processing realm learn to use Levels first. Its a wonderful diagnostic tool which reveals what was actually captured in the photo and can make changes to the overall contrast in the darker middle tones very simply. Is it the best tool in Photoshop to do that? No. But if you start with an understanding of what it does on a cause and effect level to change technical parameters in a digital capture to match perceptual expectations you will better understand how and why the more sophisticated techniques do the same thing, albeit with more precision and less "bruising of the fruit" especially if you use 16-bit ProPhotoRGB as your working space 
|
