Using 'Linear Response' curve rather than 'Film Standard' in C1 ?

[Doug Peterson]

Nice to read somebody having a deep knowledge of Capture One here. And thank you for the hint, that we may find an answer to our questions in your tutorials.

Perhaps you would be more successfull in selling them to the folks here if you gave us a demonstration of your capacity to answer the question asked about the icc profile, which would serve le linear curve as well as the default icc curve serves the standard curve. I asked that same question in Niel's blog and am still waiting for an answer.

Thanks, though the need for "demonstration" in somewhat questionable as we've never had a COMP class that wasn't sold out. I'm guessing from the fact that you have 5 posts here than you're new - so you've no doubt missed the literally hundreds of times I've answered tricky questions about Capture One (and Phase/Leaf/Hassy hardware) here and on other forums.

In this case I'd have to defer the complex, use-specific (even if you think it is not), and niche discussion of proper pairing of [ICC profile] and [camera curve] with any given [use case] to our more in-depth classes.

Someone had commented that there should be more webinars/classes for these sorts of advanced topics, so my link seemed relevant.

Thanks for the note about the website - I'll have our web person correct that.

[torger]

With a recent version of Capture One you can output TIFF files with embedded camera ICC profile and also get the TIFFTAG_TRANSFERFUNCTION set. If you reverse that transfer function you should get back to linear camera data, pre-processed with white balance but otherwise in native camera color space. If it's a Leaf ICC profile it has also been preprocessed from camera RGB to Prophoto RGB though, using the CaptProf_color_matrix embedded in the raw file.

Note that a "linear" curve will not look linear, as it must also take into account the ICC gamma, which generally is 1.8. For an Aptus-II the resulting transfer function from a linear curve is almost exactly an x^1.8 function, ie effectively linear as the embedded ICC profile has gamma 1.8. For many other cameras (almost all?) there's deviation though, typically the transfer function leaves some space in the highlights. This might be due to give the ICC cLUTs proper headroom or to hide a noisy clip level which some cameras have.

My current understanding is that the bundled ICCs are generally designed for the "standard curve" or curves similar to it - probably all curves *except* the linear curve is "similar enough" to the standard curve. A linear curve is however far from this standard curve and therefore may cause the LUT conversions become a bit distorted, and thus it would not be recommended to use the bundled ICCs when linear curve. I've tried to find some "official" statement regarding this but not received any though. I guess they want you to take a course

I have not tested that many images but with a Leaf back and a Canon 5D mark 2 one can see that (white) skin get a slight greenish cast if the ICC is applied to a linear curve while it looks good with the standard curve. Depending on subject it may be difficult to decide which one is better, but if you layer the two images on top and switch fast between them one can clearly see that the color does change despite that you're using the same ICC profile, and with this fact I think it's reasonable to assume that the bundled ICC profiles have been designed to look good with the standard curve.

[Vladimirovich]

Note that a "linear" curve will not look linear, as it must also take into account the ICC gamma, which generally is 1.8.

extract profile, replace gamma with gamma1, replace profile...

[torger]

extract profile, replace gamma with gamma1, replace profile...

An interesting experiment, not sure what it would yield though. Do you mean that the modified profile would render colors correctly with the linear curve applied?

Most of the bundled profiles actually don't have a 1.8 gamma curve in the TRC tags, it's rather a 256 point curve with a shape which is similar (but still quite different from) a 1/1.8 gamma curve, but the effective gamma with the A2B0 LUT active becomes 1.8. I'm new to LUTs and have not figured out all details with them, but I know as much that you can do lots of tricks in the LUT, I assume this includes changing gamma.

[tho_mas]

extract profile, replace gamma with gamma1, replace profile...

how do you do this? with what software? I wonder whether it's possible to change the Gamma of table based profiles just like in matrix profiles? In my opinion it's only possible with matrix profiles (essentially defined by primaries and TRC and WP)... and if you change the Gamma of an LUT profile you'll screw the entire profile (the way it models colour that is).

side note regarding my posts above: reading the other thread on this forum regarding Leaf profiles... I've looked into the Leaf profiles and these do infact contain different TRCs. All other input profiles (at least those I've ever inspected) effectively assign Gamma 1.8 ...

[Vladimirovich]

how do you do this? with what software? I wonder whether it's possible to change the Gamma of table based profiles just like in matrix profiles? In my opinion it's only possible with matrix profiles (essentially defined by primaries and TRC and WP)... and if you change the Gamma of an LUT profile you'll screw the entire profile (the way it models colour that is).

I assumed that poster was interested just in tiff w/ linear data in it w/o embedded ICC profiles messing w/ that, so get rid of LUTs then... no ? because naturally the whole point of non-matrix profiles is to make quite non linear color transforms

[tho_mas]

I assumed that poster was interested just in tiff w/ linear data in it w/o embedded ICC profiles messing w/ that, so get rid of LUTs then... no ? because naturally the whole point of non-matrix profiles is to make quite non linear color transforms

but the input profiles (camera profiles) in Capture One are desgined differently. The non-linearity of the RAW capture of any camera is linearized "under the hood".

The "camera profiles" (selectable in "Base Characteristics") are designed ... if you want so ... as camera-specific "working spaces" providing a neutral grey axis (producing a distirbution of luminance levels that is exactly Gamma 1.8 ... except for Leaf backs). This is why you can embed the camera profiles on export/processing and work the files with the "camera profile" embedded in Photoshop just like with any other working space. (Except, of course, with an input profile only containing an "A2B0" table you can't create composites ... at least you have to establish workarounds if want to do so).

What you infact can do... if you want to ...
- create a profile for instance with ProPhoto primaries and WP but with Gamma 1.0 (you can create a rudimentary version of it also in Photoshop)
- edit your captures in C1 with the generic or dedicated input profile and the linear curve applied and select your self baked Gamma 1.0 profile as target when processing

not sure that you will gain anything ... but at least this would be doable :-)

[AreBee]

Allegretto,

...the "linear" is the actual RAW file curve. The Others, "film standard" are curved renditions of the RAW

This makes sense to me - I currently am trialling C1 and noticed that one NEF file in particular very obviously is wrong when the default film curve is adopted, whereas the same file looks correct when the linear curve is adopted, and similar to how it is rendered in Nikon Capture NX2. Having said that, the file in question appears to be the odd one out of those that I have examined, as every other NEF looks reasonable with the film curve applied.

[allegretto]

Allegretto,

This makes sense to me - I currently am trialling C1 and noticed that one NEF file in particular very obviously is wrong when the default film curve is adopted, whereas the same file looks correct when the linear curve is adopted, and similar to how it is rendered in Nikon Capture NX2. Having said that, the file in question appears to be the odd one out of those that I have examined, as every other NEF looks reasonable with the film curve applied.

I suppose "Standard Film" has it's matches and mis-matches too. That's what all those sliders and histograms are about

As far as "linear" goes... don't tell anyone that or you will be WRONG, WRONG WRONG,,, I guess...

[ErikKaffehr]

Hi,

I tried some images using both linear and film curves. The "film curves" make your image to bright, so exposure needs to be reduced 1-1.5 stops. The linear curve looks right, but I feel colours are better with the "film curve".

It seems that colour handling is optimised for film curve.

Best regards
Erik

I suppose "Standard Film" has it's matches and mis-matches too. That's what all those sliders and histograms are about

As far as "linear" goes... don't tell anyone that or you will be WRONG, WRONG WRONG,,, I guess...

[tho_mas]

The "film curves" make your image to bright, so exposure needs to be reduced 1-1.5 stops.

The Standard Film Curve is supposed to be used with a conventional lighting meter... AFAIK. I've once tested exposure on a Kodak Q14 chart. I've measured the actual light (not the reflecting light) with a hand meter and set the camera/lens accordingly. My P45 matched the respective RGB values of the 18% grey patch very, very good with the Standard Film Curve applied.
Now, when you expose too much to the right, the Standard film curve will blow differentiation in bright tonal values. When the motif fits into the dynamic range of the camera and you like the look of the Standard film curve (I do... actually) then you can expose for mid grey. When the motif exceeds the dynamic range of the camera it's certainly more appropriate to expose to the highlights and use either the "Extra Shadow" film curve (and recover some highlights) or the linear curve.

[ErikKaffehr]

Hi,

The reason for ETTR is that we want minimise noise, or to be more exact, maximise signal/noise ratio. Exposing for highlights is essential to utilise the sensor maximally. It works of course only at base ISO. The whole idea is to maximise exposure. What I do is that I use RawDigger to check the the raw files for exposure. It is of course something I do in post, it is a learning experience.

The enclosed figure is a good example. Exposure was limited by the white wall on the castle, this was judged by histogram and blinking highlights.

I enclose the raw histogram and default rendering in C1 (using WB from a grey card shot in another frame) and my own processing in LR 5.3, the C1 looks good, in my eyes.

Best regards
Erik

The Standard Film Curve is supposed to be used with a conventional lighting meter... AFAIK. I've once tested exposure on a Kodak Q14 chart. I've measured the actual light (not the reflecting light) with a hand meter and set the camera/lens accordingly. My P45 matched the respective RGB values of the 18% grey patch very, very good with the Standard Film Curve applied.
Now, when you expose too much to the right, the Standard film curve will blow differentiation in bright tonal values. When the motif fits into the dynamic range of the camera and you like the look of the Standard film curve (I do... actually) then you can expose for mid grey. When the motif exceeds the dynamic range of the camera it's certainly more appropriate to expose to the highlights and use either the "Extra Shadow" film curve (and recover some highlights) or the linear curve.

[Bart_van_der_Wolf]

The Standard Film Curve is supposed to be used with a conventional lighting meter... AFAIK. I've once tested exposure on a Kodak Q14 chart. I've measured the actual light (not the reflecting light) with a hand meter and set the camera/lens accordingly. My P45 matched the respective RGB values of the 18% grey patch very, very good with the Standard Film Curve applied.

Hi,

That's because it underexposes the Raw exposure relative to it's ISO setting. The subsequent highlight clipping by using a Film response curve then corrects for that underexposure, or it clips highlights if scene contrast is high and incident light metering was used uncorrected.

Now, when you expose too much to the right, the Standard film curve will blow differentiation in bright tonal values. When the motif fits into the dynamic range of the camera and you like the look of the Standard film curve (I do... actually) then you can expose for mid grey. When the motif exceeds the dynamic range of the camera it's certainly more appropriate to expose to the highlights and use either the "Extra Shadow" film curve (and recover some highlights) or the linear curve.

The linear response curve also improves highlight contrast. Certain subject matter benefits a lot from that, e.g. white clouds in a sunny scene.

Cheers,
Bart

[tho_mas]

we want minimise noise, or to be more exact, maximise signal/noise ratio.

No. What we (photographers, imaging people, designers … whatever) want is a nice, sexy image. When the lighting of the scene fits into the dynamic range of the camera (sensor) why on earth would I "overexpose" (overexpose in the sense of ETTR when it's not needed). Just to fiddle around in software and to dial down the brightness afterwards? Ridiculous!

That's because it underexposes the Raw exposure relative to it's ISO setting. The subsequent highlight clipping by using a Film response curve then corrects for that underexposure, or it clips highlights if scene contrast is high and incident light metering was used uncorrected.

The linear response curve also improves highlight contrast. Certain subject matter benefits a lot from that, e.g. white clouds in a sunny scene.

Bart, I know all that. I do use the linear curve as well. But I use it when needed and not as the starting point of all my editing.
The non-plus P-backs show the histogram with the linear curve applied on the LCD (unlike the P plus backs that show the histo with the Standard Film Curve applied on the LCD). I use both a non plus and a plus back and know how to expose with both of them to get a decent image. But, again, when I do not clip the blacks nor the highlights in a certain scene … then I expose for a "nice" image rather than for highlights. Why would I?

[ErikKaffehr]

Hi,

I guess I need to retract a bit on the necessity on ETTR, I tried to demonstrate the effect by example. What I have seen is that the improvements I expected from ETTR were visible in histograms but not really visible in photographs. That said, I still feel that maximising exposure is a good idea.

What I also see is that histogram and blinking highlights on my P45+ are pretty good indication on where exposure will end up.

Best regards
Erik

No. What we (photographers, imaging people, designers … whatever) want is a nice, sexy image. When the lighting of the scene fits into the dynamic range of the camera (sensor) why on earth would I "overexpose" (overexpose in the sense of ETTR when it's not needed). Just to fiddle around in software and to dial down the brightness afterwards? Ridiculous!
Bart, I know all that. I do use the linear curve as well. But I use it when needed and not as the starting point of all my editing.
The non-plus P-backs show the histogram with the linear curve applied on the LCD (unlike the P plus backs that show the histo with the Standard Film Curve applied on the LCD). I use both a non plus and a plus back and know how to expose with both of them to get a decent image. But, again, when I do not clip the blacks nor the highlights in a certain scene … then I expose for a "nice" image rather than for highlights. Why would I?

[tho_mas]

That said, I still feel that maximising exposure is a good idea.

Me too as long as we talk about increased exposure to get the shadows of a high dynamic range scene into the histogram (without clipping highlights of course). But when the dynamic range of the scene can be fully captured by the camera there is no need to push exposure, IMO.
In my experience you can beat a file of a Phase back excessively in post without introducing unwanted noise or any other artefacts (tonal banding... whatever) ... provided that the capture didn't clip shadows or highlights initially. This goes even for my old P45 (non plus ... from 2005).

[Vladimirovich]

provided that the capture didn't clip shadows

and what does it mean exactly - clipping shadows in raw channels ?

[tho_mas]

and what does it mean exactly - clipping shadows in raw channels ?

yes

[Vladimirovich]

yes

I was asking what do you mean by clipping shadows when you are talking about raw ? is it somewhat related to the high readout noise of consumer level MF cameras v CCD sensors vs consumer level cameras w/ CMOS sensors ?

[ErikKaffehr]

Hi,

What I see is that the shadows don't clip on my P45+ but go into a continuum of noise. On my Alpha 900 I have seen some clipping of dark tones.

I see two reasons to maximise exposure, one is to keep down shot noise in light and midtone areas. I tried to demonstrate this and it is easy to show the issue mathematically but signal noise ratio is so good that the effect is not obvious in practice. I can measure it but I cannot see it.

The other reason is to protect shadow detail. Higher exposure gives less shot noise in the darks.

In general I would say that noise in dark areas is an area where the P45+ doesn't impress. It is also a reason that they don't really impress at higher ISOs.

Chris Barrett published some raw images from Sony Alpha 7r and IQ260 (?), which he exposed at 100 ISO. Those IQ260 images were noisy in both dark midtones and in the darks, while the Sony image was smooth. Both IQ260 and Sony Alpha have a base ISO of 50 (nominally).

What I have seen on my P45+ is that shadows are more problematic than on the Alpha 99, so I feel a greater need for exposing high on the P45+ than on the Alpha.

Best regards
Erik

I was asking what do you mean by clipping shadows when you are talking about raw ? is it somewhat related to the high readout noise of consumer level MF cameras v CCD sensors vs consumer level cameras w/ CMOS sensors ?