Colour Casts

[Colin O]

<p>I'm going to ask an obvious question: How can I tell if an image has a colour cast?<br>

(Let's assume I am using a calibrated monitor and a colour-managed workflow.)<br>

Is there some systematic way to tell, or is it just a matter of looking at an image and the colours look "wrong"?</p>

 

<p>Say for example, I scan an image and it has an obvious colour cast. I can use Photoshop's 'Colour Balance'

command to remove the colour cast. But how do I know by how much to adjust the sliders? Is it just performed "by

eye" until it looks right?<br>

What about the situation where there's only a very slight colour cast? Sometimes a colour cast only becomes

apparent when you see the image with the colour cast removed.</p>

 

<p>If you consider Photoshop's 'Variations' command: You can click to make an image "More Green", "More Yellow",

"More Red", etc. Sometimes I find that I don't know when to stop. I don't seem to be able to tell at exactly what

point the colour cast has been neutralised.<br>

Sometimes the longer I look at an image, the less I "see" the colour cast - a bit like when you watch a movie

with an intentional colour cast - after a while your brain just "auto-adjusts".<br>

Sometimes I think I have removed the colour cast, only to return to the same image another day and see the colour

cast hasn't been fully removed.<br>

Sometimes I think an image is fine - until I see a variation that looks better.</p>

 

<p>I'm interested to hear how others approach this.</p>

[JDMvW]

<p>You can do manually what "Auto Color" does automatically. Use levels and then select the lightest area of the image for the 'white' dropper, and the darkest area of the image for the black dropper.<br /> 99% of the time, this will 'correct' the image.</p>

<p>The variations command display is not intended for 'color correction' but for 'color selection', as you've demonstrated empirically.</p>

<p>Digital is better than film in color accuracy, I think, but there is really no such thing as one "correct" color balance that will match the imperfect perception of the human eye/brain combination.</p>

[digitaldog]

<p>If you <strong>know</strong> you want to define a neutral value, in RGB working spaces, that is the case when you have R=G=B (34/34/34).</p>

[tom_mann1]

IMHO, for most common photographic subjects (eg, family snaps, landscapes, art photography, street, situations where

the lighting is not under your control, etc.), there is no single best color balance. Not only are the best colors subjective,

but there are often different color casts in different areas of the image, as well as different color casts in the brights, the

mids, and the shadow areas. In fact, for the best visual / psychological effect of the image, one rarely wants true color accuracy.

 

However, for some very important technical uses of photography (eg, copying /reproduction work, advertising, etc.), it's

critically important that the colors that the final viewer sees match the original item as closely as possible. In these cases,

the original often is available for comparison. In addition, the photographer will have stable, color calibrated lights, a color

neutral studio environment, a profiled camera, etc., and even with all of these precautions, will still include a shot of something like a McBeth color checker chart anytime any one of the above things changes. Comparison of the final rendition of the color checker with its known values provides the basis for final, hopefully small, color ttweaks.

 

Tom M

[Bill C]

<blockquote>

<p>Sometimes the longer I look at an image, the less I "see" the colour cast - a bit like when you watch a movie with an intentional colour cast - after a while your brain just "auto-adjusts".</p>

</blockquote>

<p>Yep, this effect is well known in professional photo labs. The general way to deal with this is to keep various reference images around, and these can be used to rezero your internal reference point. The person making the corrections has to be aware of how easily their perception can shift, so they have to periodically look around the room or at the reference images.</p>

<p>Working strictly from a monitor is a bit trickier, and in the large lab where I spent time, if we didn't like the way someone was correcting, we'd just pull a set of prints (from well-controlled and profiled printers). They could see how the prints look in a proper light booth, compared with how the monitor images looked. Then they can sort of recalibrate their internal reference to correlate how the monitor and physical prints relate.</p>

<p>Personally, I like to use the numbers as a cross reference. In something like sRGB or Adobe RGB, neutral colors (gray) have equal values, R=G=B. For anything else, I like to use numbers in a color space known as CIELAB commonly just called "Lab." I've mentioned this a time or two on photo.net, with respect to portraiture, but so many people here seem to flat reject the notion that I don't generally bother wasting my time anymore. But with CIELAB, skintones generally fall in a fairly narrow range for the a* and b* values and this is a very good way to crosscheck your visual judgement (most of the time skintones are nowhere near as subjective as many people think). Maybe I should mention that I've worked in photography my entire life, including plenty of time in large pro labs.</p>

<p>ps; maybe I should mention that monitor images seem to have a lot more leeway than prints, with respect to how much variation is ok.</p>

[John Seaman]

<p>If there's an area of the image which should be a neutral grey, for example a tarmac road I open up the info palette and run a levels adjustment. Moving the cursor over the grey area , adjust the red, green or blue channels until the three numbers are the same. Otherwise yes, do it by eye, given the limitations on this which you have noted.</p>

[Tim_Lookingbill]

<blockquote>

<p>Say for example, <strong>I scan an image</strong> and it has an obvious colour cast. I can use Photoshop's 'Colour Balance' command to remove the colour cast. But how do I know by how much to adjust the sliders? Is it just performed "by eye" until it looks right?</p>

</blockquote>

<blockquote>

<p>Sometimes I think I have removed the colour cast, only to return to the same image another day and see the colour cast hasn't been fully removed.<br />Sometimes I think an image is fine - until I see a variation that looks better.<br /> I'm interested to hear how others approach this.</p>

</blockquote>

<p>Depending on the media you're scanning (i.e. negatives, positive film or prints) there can be some complex color table shifts exacerbated by too much/less contrast along with what simple WB can fix. Overly blue shadows and overly yellowish orange/red midrange/highlights along with too much saturation can trick the eye's ability to judge a color cast. There's two corrections more pronounced in film you have to keep in mind that can cause these optical tricks. The first being simple WB and second the individual color elements of each object in the scene.</p>

<p>I can only show you with past discussions linked below demonstrating these optical effects that make it hard to determine which of the two or both are influencing the overall color cast appearance.</p>

<p>The first discussion involves a high contrast very yellowish looking image whose contrast and thus saturation has been cranked to 11. Saturation/contrast can affect color cast just as much as what WB sliders can do causing back and forth edits to get it right.<br /> http://www.photo.net/digital-darkroom-forum/00Toub</p>

<p>This second discussion shows the odd cyan stain that can permeate throughout typical scans of negatives by software that can't effectively remove the orange mask. Go to the previous page of the linked discussion below that is a lead in to what I'm talking about. Removing this sickly cyan stain required individual color table adjusts using a Hue/Sat/Lum edit which affects what is called in color science as "Color Constancy".<br /> http://www.photo.net/digital-darkroom-forum/00MRuZ?start=50</p>

<p>Hope this gives you an understanding of the importance of calibrating and profiling a digital color capture device which attempts to remove these types of color errors in an image. Negatives can't be profiled so sometimes HSL adjusts are required to over come color cast created by software removing the orange mask.</p>