Photoshop Saturation vs Raw Therapee Colorfulness

[ned1]

<p>I thought I would post some variation on a Raw file processed in RT and PS. Each started with a neutral setting, color temperature set to 4750, with a bump up in exposure compensation.</p>

<p>I'll start with the Camera Raw output:</p>

<p> </p><div></div>

[ned1]

<p>Next the Raw Therapee output, 4500K, neutral setting except bump in exposure</p><div></div>

[ned1]

<p>OOP, sorry, both of the above images were 4750K</p>

[ned1]

<p>OK, now PS Camera Raw with a bump in saturation:</p><div></div>

[ned1]

<p>Now RT with colorfulness bumped up</p><div></div>

[ned1]

<p>Personally I detect a distinct difference between the two</p>

[ned1]

<p>And finally, here's the original RT image, opened in PS (as 16bit tiff), then given a saturation bump. In PS, as saturation increases the relative differences between the various reds also increases. In RT their relationship remains pretty consistant even if I do a big bump</p><div></div>

[ned1]

<p>Conclusion: Bumping up colorfulness in RT gives me more realistic skin tones than increasing saturation in Photoshop. I find the results I'm getting from RT to be not that far off of what I get with a high sat chrome film (like the old Astia)</p>

[ned1]

<p>Oh, the camera was a D700</p>

[ned1]

<p>I believe the explanation is that saturation bumps up colors relative to their current luminosity. Colorfulness (according to the wikipedia entry if you believe it) bumps up colors relative to grey. Thus bumping up saturation may bump up different colors by a different relative amount, but colorfulness should bump them all up in tandem relative to a common grey.

 

No the subject did not have a red nose.

[digitaldog]

<blockquote>

<p>Conclusion: Bumping up colorfulness in RT gives me more realistic skin tones than increasing saturation in Photoshop.</p>

</blockquote>

<p>Duh! That's why Photoshop (and originally Adobe Camera Raw) have Vibrance and I've already explained the history in that other post you started on the same subject. <br>

NO two raw converters, even with <strong>identical settings</strong> produce identical rendering. You can set two of them to identical CCT Kelvin values and you'll get different results. Kelvin defines a <em>range</em> of colors for one**. The values are correlated. Then there are all the proprietary processing each converter applies. And each converter has to make some broad assumptions of the native camera color space. So you're not uncovering anything earth shattering here. <br>

I can provide examples where a linear saturation bump will provide better subjective color than Vibrance or whatever you want to call the controls in your raw processor. That's why you've got options and more than one way to control saturation. </p>

<p>** http://www.ppmag.com/reviews/200512_rodneycm.pdf</p>

[ned1]

<p>vibrance is not colorfulness.</p>

[digitaldog]

<blockquote>

<p>vibrance is not colorfulness.</p>

</blockquote>

<p>Fine, doesn’t mater as was pointed out to you in another tread. What's your point? <br />Colorfulness (you got the spelling either right or wrong depending on WTF you're referring to) is a name of a slider or control in the software you seem to prefer that isn't anything like Vibrance or Saturation in another product. <strong>So what?</strong> They could call it anything they want. You move the slider and you get a color appearance. Big deal! Are you suggesting that colorfulness as it's called is something that no other product can produce? <br /> I pointed out a fact you seem to wish to ignore: all raw converters have controls, names and processes that are unique and proprietary. Again, noting earth shattering here, at least for those of us that have investigated numerous raw converters. Is there any reason you continue to take us down this path?</p>

<blockquote>

<p>Each started with a neutral setting</p>

</blockquote>

<p>Do you understand there's really no such thing and it's moot? My neural setting may be quite different from yours. The entire idea behind using a raw converter on raw data is so you have options! </p>

[photomark]

Edward, looking at the samples suggests that there is no relationship between the slider marked colorfulness in RT and the colorfulness as it's

described in the color modeling literature. Andrew is pretty much right on the money. It might be an interesting exercise to shoot a test chart and try to

figure out what is going on if this sort of thing interests you. But it will be really difficult to make any sort of meaningful comparison in a portrait

beyond subjective judgements—not that these aren't valuable.

 

The newer color models are trying to account for various visual phenomena that traditional that colorimetry doesn't predict. Things such as

simultaneous contrast—the effect a surrounding color has on perception. Traditional colorimetry will tell you that a gray surrounded by black is the

same as a gray surrounded by white. But our eyes and countless optical illusions tell us differently. If you want to pick two grays that match within

these different surrounds, color appearance models can help. This is the world that the technical definition of colorfulness fits into. It's useful when

predicting phenomena like the 'Hunt Effect' where changes in luminosity cause changes in perceived color purity.

 

Since a single image generally involves having your eyes at a single state of adaptation and a single white point, it's difficult to imagine a concept like

colorfulness that is absolutely defined ever being useful. If the brightness of the white point is fixed, chroma, saturation, and colorfulness are all going

to move in unison. You might be able to make a case for colorfulness (by this definition) if you have absolute data about the luminance in the scene

—i.e. you would could distinguish between an otherwise matching RGB color of a gray that is brightly lit and a white that is dimly lit— but you don't.

 

An area where you might use a concept like colorfulness is trying to precisely match colors viewed in dim surroundings with colors viewed in bright

surroundings or in tone mapping applications. But that's going to be significantly more complex than a slider and something you wouldn't do while

processing a single image.

 

Also, saturation, at least as implemented in photoshop is just a simple RGB transform—it doesn't do anything fancy. You can calculate it yourself on

the back of a napkin. Take your RGB triplet, subtract the lowest value from the highest value and divide the result by the highest value. It's very

primitive. For example, the saturation of the RGB color [150, 120, 100] = (150-100)/150 = 33%. Brightness and hue are similarly primitive. For

instance brightness = (highest RGB value)/256. Which is why [192, 192, 192] and [0, 0, 192] are both 75%. Changing saturation is just solving the

problem of keeping hue constant while you alter the saturation—a simple linear problem.

[Tim_Lookingbill]

<p>Edward, those are some gorgeous skin tones coming out of Raw Therapee, but I have to tell you after processing about 3000 Raw images you're not comparing apples to apples with your side by side tests.</p>

<p>First off you've applied a linearization (flattened contrast) tone map on the RT version of your image to apply your saturation boost. The same contrast influenced effect on saturation can be done in ACR/LR by changing the default settings to look more like your Raw Therapee version with regard to contrast and modeling of skin detail.</p>

<p>If the Raw Therapee version gives you better looking default contrast tonality demonstrated above I'ld suggest you stick with it because I don't get that look in ACR. I have to work each image. Maybe I should really make myself change to a better ACR default setting that renders like your RT version. </p>

[GBarrington]

<p>Well, the takeaway for me is that RawTherapee is pretty darned good.</p>

[Tim_Lookingbill]

<p>That's a very insightful response that I'm sure will help a lot of folks, Glen.</p>