Novel method of colour correction?

[rodeo_joe1]

<p>I've not seen this method documented anywhere, so I think I might have stumbled on a novel way of correcting poor colour balance by attempting to emulate the use of a colour-correcting filter on the camera.</p>

<p>The method is to sample a "neutral" area of the image that needs correcting using the colour picker, so that it becomes the foreground colour. Next add a new layer to the image and pour the selected colour into the new layer using the flood fill tool. With the flooded layer selected, use the Hue/Saturation tool to rotate the colour by 180 degrees (in other words complement the colour). Now select the Layer blend mode "Color", and adjust the opacity of the layer to remove the original colour cast.</p>

<p>While this seems a bit more trouble than using Levels or Curves, it appears to me to give a more even removal of a colour cast across highlights and shadows. The correction layer can be tweaked in real time using the Hue/Saturation tool, and it also seems to be less prone to posterisation effects, since you're actually adding colour and not removing it. Another benefit is that if some areas are overcorrected the new layer can be locally and/or partially removed using the eraser tool.</p>

<p>Comments please?</p>

[John Seaman]

<p>Interesting idea, thanks for sharing, it certainly works. I found an opacity of 30% worked best.</p>

[geoffrey_swenson]

<p>"..the new layer can be locally and/or partially removed using the eraser tool."<br>

Why not just add a layer mask? It is reversible.</p>

[patricklavoie]

<p>Unfortunately, look like a old method i use years ago.. before Adobe Camera Raw.. look similar to your, but simple as you dont have to select anything manually.. all can be achieve via a action.</p>

<p>1_double your background image</p>

<p>2_select filter / blur / average color</p>

<p>3_invert this layer (cmd+i or ctrl+i)</p>

<p>4_blend mode COLOR, adjust opacity to taste.</p>

[tom_mann1]

<p>Interesting idea, Rodeo. I don't think I've seen exactly this approach ever described in print before, either. Thanks. </p>

<p>I tried it on 3 images and compared the results to;<br>

(a) bringing the image into ACR and then using the WB eyedropper;<br>

(b) using the old "average color" method that Patrick mentioned (and is mentioned much too often in beginner tutorials);<br>

© using the "White Neutralizer" in NIK Color Efx Pro; and,<br>

(d) using some of the WB correction filters in Tiffen DFX.</p>

<p>In all cases, I much preferred the results from ACR's eyedropper. HOWEVER, in some cases, the method you described gave an interesting old-time effect. It could range from a sepia toned color print (when starting with an overly blue original) to almost an old 2 or 3-striped Technicolor look (when starting with an overly warm original). I'm sure the same look could be obtained with other methods, but it definitely is an interesting approach and one to keep in your tool kit.</p>

<p>Cheers,</p>

<p>Tom M</p>

[rodeo_joe1]

<p>Tom, I agree that if you start with a high bit-depth RAW file the easiest route is to simply correct the WB in ACR or whatever flavour RAW editor you like. I thought the described method might be less destructive on fixed 8 bit images though. And yes, the saturation does take a knocking if the colour cast is strong, but it can be brought back quite well with a simple Saturation boost.<br>

I've compared this complementary layer filter to more conventional methods, and it seems to induce less shadow noise, or rather exaggerates any existing noise a bit less.</p>

<p>Patrick, averaging the colour presumes that the underlying image will integrate to grey - not always the case, and inverting the layer will alter its brightness, so surely it's better to swing the Hue/Sat through 180. And how can you adjust the opacity to taste within an action?</p>

[patricklavoie]

it dont assume u have gray, it make a layer base on the predominante color, witch often result as a color cast. by

inverting this color you get the oposite color. and by seting it to color mode you are creating a kind of transparency filter ..

kind of the same old kodak filter u use with a colormeter back then.

 

with a action u cant really set a opacity to different result, but u can set it to 25% let say and manually adjust it to taste...

the whole process can be automate except this last step.

 

as for the brightness, it should not really change at all if u do it as i explain.. try it.

[JDMvW]

<p>One of the things I love about Photoshop is that there are almost as many ways to any given end as there are photographers.</p>

<p>Thanks for sharing.</p>

[robert_k1]

<p>Before removing a color cast, one should first decide whether the cast is desirable or not.</p>

<p>In an old thread, I commented:</p>

 

<blockquote>

<p>When I started digital color correction many years ago, I also wanted to "find" the neutral tones and apply the eye droppers. More often then not, that would send me down the wrong path. Then I read Dan Margulis' book in which he talked about what are *true* neutral tones, how to *decide* whether they really exist in an image, and finally how to correct if they do/don't exist. That small section was worth the price of the book.<br>

<br /> A "true" neutral tone may not exist in our eyes even *before* releasing the shutter, e.g. a sunset that has a golden cast on everything, or a face lit by candle lights. In cases like these, including a gray card in the captured scene will not do any good. Nor will it make sense to digitally remove the *desired* casts by using an eye dropper, etc.<br>

<br /> OTOH, when shooting in a studio under controlled lighting, including a gray card would make sense. Such as when shooting products and preserving their true colors. Then digitally applying the eye droppers on the gray card in the captured images will quickly remove the *undesirable* casts from the studio lighting.<br>

<br /> But the vast majority of my images shot under natural lighting do not have any "true" neutral tones. More often than not, the seemingly "gray" pavement, the "white" cloud, and the "black" shadow are not "true" neutral tones. Once I apply an eye dropper to any of these, and if it happens to be not "true" neutral, it would be the beginning of a wrong path.<br>

<br /> BTW, these comments also apply to white balancing.</p>

 

</blockquote>

<p>http://www.photo.net/digital-darkroom-forum/00VNcE</p>

[rodeo_joe1]

<p>Each to his own Robert. For most subjects I prefer to have as close to a truly white White Balance as possible. "Bright and White is always right, dull and grey is not the way" as some self-styled colour Guru once said. I think you'll agree that sickly skin tones from fluorescent lighting or an overall blue cast from shooting in the shade are often not very desirable, and it's amazing how a "correct" rendering makes colours pop more vividly, rather than being hidden behind the overall murk of a complementary cast.</p>

<p>Obviously you wouldn't want to correct a sunset or make candlelight look like daylight and the odd arty filter effect is fine. But only when it's wanted and intended.</p>

<p>You are actually talking to an old-hand here BTW. I started out using PhotoShop when version 3 was current back in the mid 1990s. And before that I used a nice cheap little program called Iphoto - not to be confused with i-Photo. Anyone remember it? Come to think of it, I don't know why Apple haven't had a lawsuit filed against them by the publishers of the original. It's also amazing to think that an entire image editor fitted onto two or three 1.44 MB floppy disks in those days.</p>

[john_reynolds10]

<p>Can Photoshop give true complements in terms of the colour wheel? The opposite of yellow, for example, is purple yet Photoshop using HSB gives blue. Likewise the opposite of red is green, yet HSB gives cyan.</p>

[tom_mann1]

<p>Different color models have different complementary colors. One major distinction is additive vs subtractive models, but there are other models, as well.</p>

<p>Both the RGB model and the pairs coming from traditional sources such as painting are discussed in this <a href="http://en.wikipedia.org/wiki/Complementary_colors">nice Wikipedia article</a>. </p>

<p>Tom M</p>

[tom_mann1]

<p>Oh, I just realized that I did not directly answer your question about how to obtain (in Photoshop) the set of complementary colors that you prefer. <br /> <br /> All you have to do is switch from using RGB to CMYK and, as expected, the complementary colors change from one set to a different set. To test that this works, just open two new, empty documents in PS, one in RGB, and one in CMYK. Fill both with the same color, and then hit cntl-I to invert each image. You should see the correct (but different) complements in the two color spaces.<br /> <br /> HTH,<br /> <br /> Tom M</p>

<p>PS (in edit) - I haven't actually tried what I just suggested, but it occurred to me that since there are quite a few color spaces, it's possible that even in CMYK, you won't get the complementary colors *you* want. In this case, just use the hue/saturation tool to rotate an initial color to whatever color you want. You don't always have to go 180 degrees around the circle.</p>

[john_reynolds10]

<p>I'm interested in <em><strong>precise</strong></em> complementary colours (and precise primaries, split complementaries, analagous and all manner of other colour schemes) that would print out according to conventional colour theory principles. It seems that your suggested techniques are still eyeballing. It's strange that Photoshop can't translate the colour models that you linked to in your first post to be able to do this.</p>

[tom_mann1]

<p>Hi John - I'm back at my Photoshop computer now and checked how PS handles complementary colors. In the RGB and Lab color modes, the complementary colors as defined by PS are indeed exact in that if you blend any color with it's inverse (at 50% opacity), you will get a perfect gray. However, in CMYK, as I feared, the complements certainly are not those suggested by classic color theory. They are closer to the classic complements than the RGB complements, but certainly not the same. Even worse, in CMYK, in the few cases that I tested, there was no blend of a PS defined inverse with the original color that yields an exact gray. The result you get depends on the original color as well as on which of the many variants of CMYK you are in, so, as you suggest, eyeballing seems to be required.</p>

<p>My guess is that the problem comes from the K channel, as different CMYK values (ie, 4 variables) can all translate back to the same RGB value (3 variables). I'm not sure I would consider this "strange" as much as a limitation of the underlying mathematical theory. </p>

<p>Perhaps high end digital painting programs such as Corel Painter have the ability to automatically generate the classically defined complementary colors. </p>

<p>Best regards,</p>

<p>Tom M</p>

<p>PS - Your question was one of the more interesting ones I've seen in a while on photo.net. I don't think I ever seen anyone ask it before now.</p>

<p> </p>

[Tim_Lookingbill]

<blockquote>

<p>It seems that your suggested techniques are still eyeballing.</p>

</blockquote>

<p>You're forgetting that Photoshop is color managed which uses the very "eyeballing" color connection space, Lab, that was modeled and designed from actual human perception experiments. Finding the exact compliment of another color on a transmissive viewing device may require you find color pairs that when layered using Multiply or other blend modes that offer transparency effect form a neutral gray which technically and spectrally signify as "exact" compliments. I'm guessing that whatever the number combo derived to make this happen it's not going to line up with what a spectrophotometer says.</p>

<p>This "perfect neutral gray signifier" does not happen in the physical world of ink on paper due to the spectral variances caused by impurities in both the paper and ink/pigment formulations as well as their absorption characteristics. This also applies to the physical world of CMYK which is why there are slight variances of percentages for neutral gray between all the CMYK profiles. Mix any paints and watercolor dyes from any hobby shop and try to get a neutral gray using compliments. It'll take a while I can guarantee, cuz I tried it.</p>

<p>And besides that it's not feasible to use Photoshop as an exacting scientific measuring instrument when its very design is built for human perception based editing which varies according to color relationships next each other as in a photo.</p>

[tom_mann1]

<p>John, after some reflection, I thought of a couple of ways one could generate what you want within photoshop. For example, one could define one or more gradient maps to automatically give the exact complement of any color in those color models that interest you. </p>

<p>However, it just dawned on me that what you might be looking for is a more full featured application to do this. For example, my 2 or 3 year old Color Munki came with a program called "Photo Color Picker" which can be used to automatically generate exact monochromatic, analogous, split complement, triadic, and straight complement color schemes (plus many variations of each). I attached a screen shot from that program which suggests its capabilities. </p>

<p>In addition, once I realized what you may be looking for, a quick search turned up several on-line tools to perform similar functions, ie, EXACT definition of complements and other relations between colors. These two websites caught my eye: http://www.colorsontheweb.com/colorwizard.asp , http://websitetips.com/colortools/sitepro/ .</p>

<p>Finally, with respect to the issue of the traditional color complements not being built into PS, you might find the following passage from http://en.wikipedia.org/wiki/Color_wheel informative. Basically, as Imentioned earlier, once one is considering paints, inks, and other subtractive type media, there are different definitions for a set of color complements depending on exactly which media you are using. The exact quote follows:</p>

<p><em>"...Color wheels and paint color mixing</em><br>

<em> </em><br>

<em>There is no straight-line relationship between colors mixed in pigment, which vary from medium to medium. With a psychophysical color circle, however, the resulting hue of any mixture of two colored light sources can be determined simply by the relative brightness and wavelength of the two lights.<sup id="cite_ref-Schiffman_10-1" ><a href="http://en.wikipedia.org/wiki/Color_wheel#cite_note-Schiffman-10">[11]</a></sup> </em><br>

<em> </em><br>

<em>A similar calculation cannot be performed with two paints. As such, a painter's color wheel is indicative rather than predictive, being used to compare existing colors rather than calculate exact colors of mixtures. Because of differences relating to the medium, different color wheels can be created according to the type of paint or other medium used, and many artists make their own individual color wheels. These often contain only blocks of color rather than the gradation between tones that is characteristic of the color circle..."</em></p>

<p>HTH,</p>

<p>Tom M</p><div></div>

[Tim_Lookingbill]

<p>So what does this scientifically finding compliments have to do with color balancing a photo?</p>

<p>In fact I tried the technique mentioned by the OP and consistently got undesirable results that required more work than it was worth.</p>

<p>ACR's Color Temp slider on all my images including both jpegs and Raw made quicker work of it. </p>

<p>I also find it strange not a single contributor including the OP didn't post any before and after image samples to show that this method actually works better on a wide range of images shot under varying WB lighting conditions than just doing this in ACR/LR. </p>

[tom_mann1]

<p>Hi Tim - There was thread drift. John brought up color complements after it looked like the original subject had run its course. Of course there is no relationship between color schemes and white balancing. I probably should have suggested starting a new thread on the new topic, but got lazy and didn't bother to do so. </p>

<p>And yes, starting with my 1st post in this thread where I proposed using ACR, I think the general consensus has been that ACR / LR is indeed better.</p>

<p>Cheers,</p>

<p>Tom M</p>