WTF is "Linear Contrast" (PSCS 6)

[adan]

Advertisement (gone after registration)

This is mostly a trivia question bugging me. (Keep in mind I've been using Photoshop professionally since 1993).

In PSCS 6, in the Curves adjustment box. there is a preset called "Linear Contrast."

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

The funny thing is that it isn't "linear" - it is a very subtle S-curve to increase contrast very slightly. Which is fine - it's easy to see what it does, both on the curves chart and in the image. (see above).

But why is it called "Linear?" Was it actually supposed to linearize (or de-linearize) something or other, back in the deep dark past of digital image processing? Was it supposed to correct for some legacy non-linearized camera problem years ago? Does it have a special meaning if one is doing video tone editing (or some other kind of specialized work) with PS?

Why isn't it called "Very Slight Contrast Increase" - which is more descriptive of what it actually does.

In Adobe Camera Raw (the raw-import module - separate place in the workflow) there is also a "Linear Contrast" curves option, and that really is linear - a straight 45° line. No change to contrast, no handles.

Any other long-time PS users who remember if this once had some special "Linear" function to justify the name?

[jaapv]

IIRC -but that was a Scott Kelby presentation I saw more than a decade a go so memory may fail me- it was said to match the linearity of the contrast curve of digital photographs to the (non?)linearity of a film curve. Not very convincing.To make the contrast of a digital image closer to the film look one must lift the midtones and then apply "linear contrast" At any rate, this setting hasn't changed since the very early days and is still there. One would have to ask somebody like Thomas Knoll.

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[adan]

Thanks - as my question suggests, that is about what I suspected.

[pgk]

3 hours ago, adan said:

In PSCS 6, in the Curves adjustment box. there is a preset called "Linear Contrast."

In Adobe Camera Raw (the raw-import module - separate place in the workflow) there is also a "Linear Contrast" curves option, and that really is linear - a straight 45° line. No change to contrast, no handles.

The weird thing is that the Linear Contrast curve in the RAW converter is a 45 degree line whereas the  Linear Contrast curve in PS6 itself is obviously darkening highlights and brightening shadows and is a slight curve as you say. I'm with jaapv on this. As the curve in PS appears to marginally lighten shadow detail and marginally darken highlights it appears to mimic the non-linearity of film where both are subject to 'reciprocity failure'. That said Photoshop is software and its naming conventions are probably from software engineers rather than photographers so vagaries probably exist because of this?

[Keith (M)]

4 hours ago, pgk said:

The weird thing is that the Linear Contrast curve in the RAW converter is a 45 degree line 

Which is also what it is in LR Classic.

[NigelG]

Could it be the case that the "contrast" of the title refers to the statistical term not the "photographic" one....?

"A contrast is defined as the sum of each group mean multiplied by a coefficient for each group" and one subset of contrasts is called a "linear contrast".

Also AFAIK  although linear can often imply "straight line" it more properly means "in a line form" so it can be a curve

 

[adan]

Advertisement (gone after registration)

You're trying to use colloquial English in a mathematics situation.

Quote

Linear functions are those whose graph is a straight line.

http://www.columbia.edu/itc/sipa/math/linear.html

Anyway (all of this is a rather broad generalization - it would take a physicist with a minor in chemistry several pages to be more precise)

Digital, in general, produces a native exposure "curve" that is linear. A pixel that captures 1 photon outputs 1 electron, and 2 photons produce 2 electrons, and 20000 photons knock loose 20000 electrons. And the Analog/Digital converter just counts the electrons (as electric charge, with a CCD, or voltage, with a CMOS sensor) from each pixel and outputs that number as a proportional brightness value. 0, 1, 2, 3...... 63, 187, 255. (or in a 16-bit situation, up to 64000±).

Film is not linear - i.e. 20 photons hitting a silver-halide grain/crystal do not produce twice as much silver (density) as 10 photons. Due to several characteristics. There is a triggering floor (generally theorized to be about 4 photons per grain or crystal) below which no silver is formed at all. This is what causes "reciprocity failure" - 40 photons arriving over 4 minutes (a time exposure) won't have the same effect as 40 photons arriving in 1 second, or 1/250th second.

And once a silver-halide grain is triggered, it all converts to silver - any additional photons hitting that particular grain are "wasted" and not recorded as more exposure.

And silver-halide crystals in gelatin are stacked - a crystal near the surface may "shade" a crystal below it from exposure.

In addition to that, chemical developers, in developing the silver, produce reaction by-products that restrain or kill the development reaction (notably bromine ions thrown off by the silver bromide crystals as they convert to pure silver). The more exposed silver to be developed, the faster the developer loses the ability to develop it (absent agitation to move developer around from highlight to shadow areas and vice versa)

This results in most film types having a "characteristic curve" that is S-shaped, with a toe in the shadows and a shoulder in the highlights that are distinctly non-linear.

https://www.licor.com/bio/guide/westerns/film

But, ideally, with an approximately linear "straight-line" portion in the middle tones. In the mid-tones, statistical probability over millions of grains and millions of photons means there is a reasonably linear relationship between photons in and silver density out (luckily for photographers!).

Now, digital camera makers have become increasingly sophisticated in adding a tone curve (a mathematical distortion of the linear sensor output, enroute to the final .dng or .jpg on the SD card) that simulates film's S-curve - the non-linear "look" of which has been burned into public consciousness over 180 years of looking at silver photographs.

But 15-20 years ago, and especially when .jpeg ruled, I can see where the PhotoShop creators saw a need to provide their own one-click "looks more like film" preset.

 

[jaapv]

Apart from that, I quite like using linear contrast, followed by a slight lift of the midtones. Much more subtle than using Levels.

[pgk]

1 hour ago, jaapv said:

Apart from that, I quite like using linear contrast, followed by a slight lift of the midtones. Much more subtle than using Levels.

I frequently use the 'Shadows/Highlights" (Image>Adjustments>Shadows/Highlights) as its quite controllable if the  'Show more options" box is checked. This allows flexibility in the relevant areas of the 'curve' and often suits my workflow well. I find levels far too coarse myself.

Edited February 14, 2020 by pgk

[jaapv]

I do as well, but mostly when I am trying to control the dynamic range. I often use contrast tools in the L channel, adjusting the colour saturation later, mostly back in RGB.

[frame-it]

On 2/13/2020 at 2:52 PM, jaapv said:

IIRC -but that was a Scott Kelby presentation I saw more than a decade a go so memory may fail me- it was said to match the linearity of the contrast curve of digital photographs to the (non?)linearity of a film curve. Not very convincing.To make the contrast of a digital image closer to the film look one must lift the midtones and then apply "linear contrast" At any rate, this setting hasn't changed since the very early days and is still there. One would have to ask somebody like Thomas Knoll.

https://www.cambridgeincolour.com/tutorials/photoshop-curves.htm

[jaapv]

That is a good link - I can only second it. I would like to emphasize the last remark (see my post #10):

Quote

Extreme levels adjustments in the RGB channel should be avoided; for such cases perform curves using the lightness channel in an adjustment layer or LAB mode to avoid significant changes in hue and saturation.

I would extend this to include a fair proportion of moderate changes.
Their advice to work in 16 bits needs strong support. For all images:

Edit:  PS has the option to use an automatic adjustment layer with virtually all adjustments, including curves, to make your editing non-destructive. Stack the adjustments, you can always go back to a previous one and correct it, go back to the top to check, and/or if desired adjust  another one and merge down visible for your end result. If you like it you can flatten. Jumping around in the history can also work.

[frame-it]

i really like their site..been reading it for some time now

 

https://www.cambridgeincolour.com/

 

[jaapv]

Yes,Ilike them as well They have been around for ages, I think well over fifteen years 

[NigelG]

21 hours ago, adan said:

You're trying to use colloquial English in a mathematics situation.

Adan - thanks for the explanation as to how the "film" tone curve shape arises from the actual chemistry - which even I could follow and was very helpful ...😀  

I was merely suggesting before that both the title term "Linear" and "Contrast" could be defined differently, and actually while  "in a mathematics situation" (see below from Merriam-Webster). In statistics data points that are not randomised but fall along a line are "linear" whether that line is straight/curved/parabolic etc (but then a straight line is obviously simply a special-case curve anyway.......)

Definition of statistics

1: a branch of mathematics dealing with the collection, analysis, interpretation, and presentation of masses of numerical data 

 

[Jeff S]

4 hours ago, jaapv said:

Yes,Ilike them as well They have been around for ages, I think well over fifteen years 

You might have linked them here about 10 years ago.  I’ve liked them since.

Jeff