Sony A7RII Dull Files?

[AreBee]

Bart,

Fascinating indeed. Thanks again.

[Jack Hogan]

My understanding is that (as Jack noted) digital cameras record light level linearly, and hence tonal relationships as they exist at the time of capture. It is also my understanding that the human eye/brain sees nonlinearly, and that a curve is therefore applied to the linear data during initial RAW conversion in order for the image to be perceived as correct when displayed.

Hi Rob, this is a common misunderstanding.

The relative luminance - whether from the scene, a print or a monitor - is ideally linear.  No matter the source, the human visual system responds non-linearly to it.  So the raw data contains a good (linear) representation of the (linear) visual information from the scene - and in theory needs no 'curve' to be presented to the eyes so that it will be perceived properly.

Jack

[AreBee]

Jack,

...this is a common misunderstanding.

The relative luminance - whether from the scene, a print or a monitor - is ideally linear.  No matter the source, the human visual system responds non-linearly to it.  So the raw data contains a good (linear) representation of the (linear) visual information from the scene - and in theory needs no 'curve' to be presented to the eyes so that it will be perceived properly.

That's odd (not your post, which seems logical): I'm sure I read that a curve is applied to RAW data (if memory serves, especially with respect to yellow hues?) so that the image is perceived as correct.

I'll investigate the text I read and get back to you.

...in theory...

And in reality?

[Bart_van_der_Wolf]

And in reality?

Rob,

What Jack is referring to is the Gamma pre-compensation that is applied to the file, and which is reversed by the display/profile. The net effect is something linear-ish.

But Human vision is complex, our eyes accommodate depending on the average luminance in a small angle of maybe 1 degree, so we are constantly applying tone curve adjustments as we scan a scene with our eyes. When we scan a print with our eyes, a print which usually has much lower dynamic range that real life, our eyes do not need to accommodate as much, so we need to do it for our eyes in the print itself.

Cheers,
Bart

[AreBee]

Jack,

Earlier I wrote:

I'm sure I read that a curve is applied to RAW data (if memory serves, especially with respect to yellow hues?) so that the image is perceived as correct.

I'll investigate the text I read and get back to you.

The text I referred to, from Real World Nikon Capture NX2 by Ben Long, p149, is as follows:

Our eyes are much more sensitive to changes in illumination at very low levels and very high levels than they are to changes in moderate brightness. For this reason, we don't perceive changes in light in the same linear way that a digital sensor does.

To compensate for this, a camera applies a mathematical curve to the image data to make brightness levels in the image match what the eye would see. This gamma correction process redistributes the tones in an image so that thers more contrast at the extreme ends of the tonal spectrum. A gamma-corrected image has more subtle change in its darkest and lightest tones than does an uncorrected, linear image.

Emphasis not mine.

Comparison images of sunflowers follows the text, one uncorrected (linear); the other (gamma) corrected.


Bart,

What Jack is referring to is the Gamma pre-compensation that is applied to the file...

Understood.

...and which is reversed by the display/profile. The net effect is something linear-ish.

Why apply a curve, only for it to be reversed at a later stage?

[Jack Hogan]

The text I referred to, from Real World Nikon Capture NX2 by Ben Long, p149, is as follows:

Our eyes are much more sensitive to changes in illumination at very low levels and very high levels than they are to changes in moderate brightness. For this reason, we don't perceive changes in light in the same linear way that a digital sensor does.

To compensate for this, a camera applies a mathematical curve to the image data to make brightness levels in the image match what the eye would see. This gamma correction process redistributes the tones in an image so that thers more contrast at the extreme ends of the tonal spectrum. A gamma-corrected image has more subtle change in its darkest and lightest tones than does an uncorrected, linear image.

Hi Bob,

Ben gets it wrong when he says 'To compensate for this...': the HVS's response to luminance is not the reason why gamma is applied.  Today gamma is applied because it more efficiently stores visual information in 8 bit files and for legacy reasons.  There is a thread around here that explains all that.

Jack

[AreBee]

Thanks Jack, and also for your earlier correction to what I wrote in a previous post.

...Bob...

Hmm.

[Jack Hogan]

I don't know if raw conversions needs to utilise the full DR of a camera. I would say that the defaults of the raw conversion should yield a decent image. The image should be tweaked for optimal looks. Having the ability to adjust the image to taste is in my view the idea of a raw based workflow.

Hi Erik,

Sure. If you capture a high-DR scene with a high-DR camera set up for maximum IQ most raw converters don't yet do a good job of it imho, mostly because they are not smart enough to figure out what portion of the 14 stop scene YOU would like to see in their 8 stop output.  Auto ADL/DRO etc. get you closer.  Advanced tools like Topaz Clarity help to squeeze in as much of the good stuff as possible in a mostly pleasing way (the HDR look notwithstanding)  - you typically still need to tweak.    But don't blame the camera if your normally processed, Max-IQ-captured, 14-bit scene DR in the raw data at first looks flat on an 8-bit/contrast ratio monitor - it would probably look much, much better on a 14-stop contrast ratio monitor with properly deep blacks.  And whoever said that an accurate image was a good-looking image anyways?  Make it eye-popping if that's what you like, without blaming the camera. 

Jack

Jack

[torger]

Adobe's bundled profile for A7r-II and also for say Pentax 645z (which has virtually identical color response, could be the exact same CFA) is deliberately desaturated in look.

Sony's sensors have strongly overlapping color filters, and separating them (=increasing saturation) means quite aggressive matrix weights with strong negative factors which make the profile less robust in extreme ranges (triggered by some artificial lights and such).

There are other ways though, I've generated profiles both for the A7r-II and the Pentax 645z and as long as you limit the blue channel a bit (render blue colors light) you can still get pretty robust behavior despite having realistic saturation.

[torger]

You can find my profiles in this post by the way:
http://forum.luminous-landscape.com/index.php?topic=104195.msg862840#msg862840

[torger]

I just recently updated the profiles for any interested. Just some minor changes due to updates in the DCamProf camera profiling software:

http://forum.luminous-landscape.com/index.php?topic=104195.msg867280#msg867280