This content has been marked as final. Show 18 replies
Hening, the Tone Curve of a profile is not intended to be perceptually linear.
Perceptual linearity (in terms of tonal values) is handled automatically by the color management system, within the limits of the color appearance models used to produce ICC profiles.
If you want "no pleasant tweaks added" then set the base tone curve to Linear (3rd option) and then set the Blacks slider to zero in Camera Raw (leave everything else, such as Brightness, Contrast, etc.) at defaults. The resulting image may be dark (most cameras tend to underexpose), so you'll have to increase Exposure a bit. And keep in mind the resulting image will probably look not-so-great unless the original scene had a dynamic range and luminance level similar to that of your display.
thank you for a VERY fast reply! -
I learn that perceptual linearity is handled by ICC profiles, not by tone curves in the PE or in ACR. -
The advice you give to avoid pleasant tweaks is the recipe for a linear image, as you have given it to Brent Townshend
This was not what I had in mind. I am aware of that a linear image will not look natural.
In terms of the scene referred workflow as described by Tindemans
what I search for might be described as a subdivision of step 2 (called creative processing). There must be an area between linear and purely arbitrary, we might call it natural. The fact that this area can not be easily defined let alone measured objectively ought not to preclude us from at least identifying and naming it. As it is now, we can not even in terms separate between natural, and beautified ad libitum. Arbitrariness is being presented as unavoidable.
So my question aimed at: (How) can I make a base curve that is natural, but not beautified (by exaggerating the contrast)?
"The Chart Wizard is designed to produce a reasonably accurate *scene-referred* color profile" - but how is the "rendering intent" of the Base Curve?
I tried your advice with the linear curve in the PE + lightening in ACR on both the ColorChecker image and a real scene. The CC could be made look reasonable. But this is an image where the histogram has headroom on either side. So I experienced it like you said: this may work with an image whose DR matches that of the medium. With the real scene, whose histogram stretched end-to-end in the first place, using the Exposure slider led to highlight clipping at once, as exspected; instead, I had to use the Brightness slider - that is increase the contrast, if I understand it correctly.
So it looks like-:
1-The task of the Base Curve is to map the DR of an image to that of the screen
2-If there is any answer to my question, it is image-dependent. On the other hand: the PE uses ONE Base Curve. So the question may be allowed: What is the "rendering intent" of that curve?
Kind regards - Hening.
I realize this is an old thread, but I'ld like to shed some light on the "arbitrariness" of acquiring and displaying TRUE LINEAR untouched sensor data as its viewed in raw converters. By this I mean making the preview display the raw sensor data without ANY tone curve or processing applied, not even the curve within the base profile.
Not all converters offer this option but those that do will show a very dark preview with its histogram data bunched up to the far left. This was what I thought should be the level you had to get to to establish a consistency by getting to and displaying the purist form of the untouched sensor data.
But from my experience with three brands of raw converters, two based on DCRAW, even this linear state is arbitrary as well because I got three different previews and histograms with the same raw file.
It seems even linear sensor data is interpretive as well.
The linear sensor data is absolute, but how you display it on the screen is up to the application. Arguably the correct way to display that linear data in an application like Photoshop is to assign a working RGB space that has a linear gamma (you can do this in Color Settings; set a custom gamma of 1.0). Most of the time, folks are simply loading a linear image into PS but then arbitrarily assigning their usual working space, which is Adobe RGB or ProPhoto RGB (gamma of 2.2 and 1.8, respectively) and __THAT__ is what is causing the preview to look very dark.
How is it possible to load an image into Photoshop with a 1.0 gamma? Once you export from LR or open from ACR with the ProPhoto space specified, it's already being set to a gamma of 1.8 isn't it? Once in PS, as you say, you can assign it a ProPhoto profile with a gamma of 1.0 but will that 1.8 to 1.0 change have any negative effects on the image?
True. However, you can set up a a Custom RGB with gamma = 1.0 as your working space. You can then convert whatever the ACR-produced image is (whether it be Adobe RGB or ProPhoto RGB) to your working space, and hence it will have gamma 1.0.
The same is true in general if you opened an image that was originally tagged as Adobe RGB (gamma 2.2) then converted it to ProPhoto RGB; the final result would then be gamma 1.8 encoded.
Furthermore, some raw converters (such as dcraw) allow the creation of image data without gamma encoding; so it's possible to open that in Photoshop and it'll look way too dark if one assigns Adobe RGB or ProPhoto RGB to it. If you instead assign a custom RGB with gamma 1.0 then it'll look normal.
I'm talking about viewing raw files with no processing or tone curve applied of any kind in raw converters, not in PS or other gamma encoded imaging apps.
The previews display very dark with histogram data bunched up to the left. Several raw converters that offer this option (DCRAW based) showed three slightly but noticeably different previews with regards to color and contrast and histogram display.
Assigning the same 1.0 gamma profile in PS produced three different previews of the same raw file from three raw converters.
I was under the impression raw linear data should render the same among all converters since it's straight off the sensor with no processing applied. It's not.
You were under a huge misconception.
A raw file without conversion is a very, very dark linear grayscale representation of the sensor's output. It does not contain any information that any human being would interpret as color.
It's only during the demosaicing and conversion that any conversion software interprets the colors based on the array of the color filters for each pixel and its neighbors according to what each converter deems to be the most desirable rendering, its own interpretation.
There is no such thing as an absolute, colorimetrically correct rendering of a raw file.
Given the considerable amount of material available on the subject by now, it's astonishing how many users still don't grasp this.
The raw papers on the Adobe site are mandatory reading, and I consider Jeff's book essential.
I'm already aware of the grayscale representation of sensor output. Even at this stage no one can prove/determine if the each grayscale pixel value recorded is actual or absolute because of the unknowable things like the camera's electronics and spectral qualities of the bayer filters.
My point in my explanation here (if you would've read more carefully) has always been about linear data AFTER DEMOSAICING WHICH IS AND CAN BE PROVEN TO BE INTERPRETIVE.
It's not the amount of info on this subject that's the problem, it's the amount of clarity and transparency that's missing. Point A to point B, connect the dots kind of explanation is required which tech people including myself fail to grasp.
Programers come out of the woodwork with their own raw converter and get you to think they've got something new and better when what they really offer is an interpretation on the same old stuff and leave out a lot of detail as to what's missing.
I've already checked out Jeff's book at the library. Most of the info I needed to know I already knew, just don't have the resources to put the rest of it into practice.
>AFTER DEMOSAICING WHICH IS AND CAN BE PROVEN TO BE INTERPRETIVE.
That's what I and everyone else have been telling you, without the yelling. Duh!
Whether anything is "linear" after demosaicing is another discussion.
I don't know where you've been in this discussion, Ramon, but it's not been about what you've been telling me but what the OP refers to as the arbitrary nature of raw processing. Their seems to be the understanding that a digital camera can be treated like precision calibrated high end scanner with repeatable results every time.
And from the way you address people on your infinite knowledge on the subject, it's clear from your many posts here in the past 6 years, you piss more people off than enlighten.
Hell, you even piss off Thomas Knoll. That's a new one for me.
If you want to get personal, remember the saying about people who live in glass houses.
Sorry I don't remember pissing TK off, but I do remember you outright threatening Jeff Schewe and me if we ever get to Texas. Jeff even told you where and when you could find him in your desolate state within a matter of days after the incident.
>but it's not been about what you've been telling me but what the OP refers to as the arbitrary nature of raw processing.
The unavoidable, necessary and obvious arbitrariness of raw conversion is precisely what I've been addressing. [EDIT:] specifically this sentence in your post #7 "I was under the impression raw linear data should render the same among all converters since it's straight off the sensor with no processing applied." [/EDIT] Your failure to grasp that is not my problem.
You seem to have a very selective memory.
I wasn't threatening to do bodily harm to anyone. I was informing you two that there are consequences to addressing people on the web in a demeaning manner. I only said that I would embarrass both of you if you treated me in public the way you two treat others and myself on the web. You don't know how many crazy people there are on the web. It doesn't hurt to be nice.
Now that I've corrected your memory on the matter...
You are right about my impression of linear data. I must have missed in all the discussions on the subject you and others informing of this. My memory tends to fail me when I'm more pissed off than informed.
Interesting that you should bring up "bodily harm" when I did not.
It's a clarification to your statement that I threatened someone as read in your previous post. I have a long history of never having that word associated with my character and behavior in all of my web forum postings in the past 6 years and in my daily life.
However, I've read many times of you wanting to take a baseball bat after people for not following color management policies.
LOL ! :D
For the sake of fleshing out the picture: I do advocate beating up morons that hand you untagged image files with a baseball bat. My well known stance has resulted in my running across remarkably few untagged image files. B)
(Pssst! It's a figure of speech, obviously. :))