A rant about HDR (my processing: Not in general)

[Manfred M]

The effect is natural if you can not guess that the image is an HDR image. Tone mapping does not matter here, cameras also use filters, profiles, etc. Different software uses different algorithms and generates different results.

I understand where you are coming from, but don't agree 100%. I find that when I use HDR, there are always some elements of the image that I don't like and I have to go back rework; contrast adjustment, dodging and burning are the tools I use most often.

That being said, with modern cameras and their high dynamic range (14+ eV for some Nikon and Sony models) I find that HDR is not really a tool that I need as much as I used to.

[Pebal]

That being said, with modern cameras and their high dynamic range (14+ eV for some Nikon and Sony models) I find that HDR is not really a tool that I need as much as I used to.

You can register 14+ EV on single shoot, but you will not be able to print it or show on the monitor. You have to map the tones in the software.

[DanK]

I have never tried SNS-HDR. I've only tried 3, I think, and my initial experiments led me to use exposure fusion instead.

As I mentioned in an earlier post, LR's HDR merge doesn't seem to distort colors much, unlike the other two packages I tried. However, it's worth bearing in mind that HDR and exposure fusion are at their core different approaches. Most HDR packages use tone mapping. If I understand correctly, exposure fusion doesn't; it doesn't alter the attributes of any pixels. It simply selects pixels from among the stacked images that are within certain exposure ranges. It's analogous to what you could get with manual blending of layers, except that the software algorithm rather than the user controls the blending.

I said "most" because I have been told by others that at least some HDR software packages include an exposure fusion option.

I have used Enfuse quite a bit (via the LR plugin), and I have generally found the results very natural in terms of color. However, the distribution of luminance is not necessarily what I want (which is to be expected), so tonality adjustments are often necessary. This has been even more true, in my experience, with LR's HDR merge.

At the risk of boring people with VERY long memories, I'll post below the pair of images that convinced me to switch to exposure fusion. First, an HDR image from PS (which at that time used HDR Pro):



The slightly psychedelic sky is obvious. What may be less obvious is that the reddish tone of the granite in the foreground is way off.

Second, the same stack processed using the default settings of LR Enfuse:



The colors are spot on.

[Manfred M]

You can register 14+ EV on single shoot, but you will not be able to print it or show on the monitor. You have to map the tones in the software.

Agreed, to a point. No computer screen is capable of showing all the colours that can be captured on a modern digital camera. The best screens come very, very close to showing the full AdobeRGB colour space and very high end photo printers approach the range of the ProPhoto colour space.

That's what the rendering intent in our screen and printer drivers is all about; translating the tones that cannot be reproduced by our devices into something that can.

What the 14+ eV does allow us to do is to take that data and turn it into an image without having to use HDRI. Some shots, like the ones Trevor posted here might exceed the dynamic range of modern high end cameras. If we blow out the highlights and crush shadow detail in a single image, we need to resort to some technique that uses bracketed exposures to capture those highlights or shadow; then some other technique, whether that be HDR, fusion or manual blending will be required to construct an image.

[xpatUSA]

Originally Posted by Pebal You can register 14+ EV on single shoot, but you will not be able to print it or show on the monitor. You have to map the tones in the software.

Agreed, to a point. No computer screen is capable of showing all the colours that can be captured on a modern digital camera. The best screens come very, very close to showing the full AdobeRGB colour space and very high end photo printers approach the range of the ProPhoto colour space.
<>

I'm not sure that @Pebal was talking about color accuracy or gamut, Manfred.

So I'm wondering how gamut and DR are related - especially when CiC's tutorial on DR doesn't mention gamut at all?

[Manfred M]

I'm not sure that @Pebal was talking about color accuracy or gamut, Manfred.

So I'm wondering how gamut and DR are related?

I'm not quite sure either, but he was responding to a comment I made about dynamic range. The moment one writes about displays and prints, the gamut and colour space come into the equation as well, as I assume that this is what he is writing about with his comments. Ditto contrast ratio.

I can't quite recall, but seem to remember that a print could handle a dynamic range of around 6-8 eV and a modern computer screen was in the 10-12 eV range.

[xpatUSA]

I'm not quite sure either, but he was responding to a comment I made about dynamic range. The moment one writes about displays and prints, the gamut and colour space come into the equation as well, as I assume that this is what he is writing about with his comments. Ditto contrast ratio.

DR is normally based on luminance (lux or cd/m^2) which, in turn, is based on the CIE human eye's spectral response. So, yes, output-referred DR can be affected if gamut-clipping occurs for some reason.

I can't quite recall, but seem to remember that a print could handle a dynamic range of around 6-8 eV and a modern computer screen was in the 10-12 eV range.

For computer screens that is indeed true.

Personally, I set my monitor brightness down quite a bit which lowers it's DR. The spec says max white is 350cd/m2 with a 10EV contrast ratio - but I set it down by about 1-1/2EV in my low-lit room, leaving a viewing DR of about 8EV.

It's a good job my cameras are barely able to capture 9EV and that their raw files usually come out less than that with my less-than-perfect exposures. Add in the well-known Foveon noise and blotching during conversion to RGB then something like 6 or 7 final EV is quite good in my world!

[Manfred M]

DR is normally based on luminance (lux or cd/m^2) which, in turn, is based on the CIE human eye's spectral response. So, yes, output-referred DR can be affected if gamut-clipping occurs for some reason.

Understanding that lux is an incident light reading based on the light falling on the subject whereas cd/m^2 (a.k.a. nit) is a transmitted value from say, a computer screen

For computer screens that is indeed true.

Personally, I set my monitor brightness down quite a bit which lowers it's DR. The spec says max white is 350cd/m2 with a 10EV contrast ratio - but I set it down by about 1-1/2EV in my low-lit room, leaving a viewing DR of about 8EV.

It's a good job my cameras are barely able to capture 9EV and that their raw files usually come out less than that with my less-than-perfect exposures. Add in the well-known Foveon noise and blotching during conversion to RGB then something like 6 or 7 final EV is quite good in my world!

Generally the recommended setting for computer screens used in image editing will fall in the 100 - 120 cd/m^2 range. The 120 cd/m^2 is the default setting for my i1 Display Pro. The room should be on the dark side; below 70 lux at the workstation. My work room is around 50 lux.

[xpatUSA]

Understanding that lux is an incident light reading based on the light falling on the subject whereas cd/m^2 (a.k.a. nit) is a transmitted value from say, a computer screen

Quite so. I should have said "illuminance" for lux.

Generally the recommended setting for computer screens used in image editing will fall in the 100 - 120 cd/m^2 range. The 120 cd/m^2 is the default setting for my i1 Display Pro.

What is the black level for your no doubt excellent screen? If we knew that, we could calculate your screen viewing DR in EV from log2(120/?).

I just placed my lux-meter on my screen and got 163 for white and 2 for black. No need to convert to luminance because a) the conversion factor cancels out in the EV formula and b) it was close enough to ignore the inverse square law.

I got 6+ EV. Probably would have got more if I had shrouded the sensor for the black level measurement.

The room should be on the dark side; below 70 lux at the workstation. My work room is around 50 lux.

Mine is quite variable because it's the living area but is definitely less than 70 lux where the monitor is.

[Manfred M]

What is the black level for your no doubt excellent screen? If we knew that, we could calculate your screen viewing DR in EV from log2(120/?).

I don't know Ted. Here are the published specs.

http://www.benq.ca/product/monitor/s...pecifications/

[xpatUSA]

I don't know Ted. Here are the published specs.

http://www.benq.ca/product/monitor/s...pecifications/

Same as mine, Manfred, except for "dynamic contrast" which we can ignore and the wider Adobe gamut of course - it says:

Brightness ( typ.)‎	350 cd/㎡‎
Contrast ( typ. )‎	1000:1‎

So the blackest you can get (theoretically) is 350/1000 = 0.35 cd/m2. You have set your whiteness to 120 cd/m2, so your viewing DR is log2(120/0.35) = 8.4EV.