Page 1 of 1

How do you bracket your exposures?

PostPosted: Wed May 06, 2015 6:40 am
by hjulenissen
http://forum.luminous-landscape.com/ind ... c=59312.30
Guillermo Lujik about exposure bracketing for HDR:
*Your least exposed shot preserves the highlights of interest (ideally highlights should be just about to start clipping = ETTR)
*Your most exposed shot produced a sufficiently high exposure in the deep shadows of interest to get acceptable noise there with your camera
*The EV interval between your shots can be properly handled by your fusion software (with ZN this gap can be 3 or even 4 stops)


He an others propose to have quite large exposure intervals (e.g. 3-4 stops), meaning that most interesting scenes can be properly captured by few brackets. The available camera DR will be limited to some degree by lens glare anyways (a post by Jim Kasson)

http://www.dpreview.com/forums/post/55285796
"Based on simulations, the Canon 20D can record nearly 20 stops of dynamic range using HDR imaging if only a point light source is present. If half of the field of view is covered by an extended source, then only 9 stops of dynamic range can be recorded by the 20D..."


So if todays best cameras offer a DR of ~14 stop at base ISO, lenses (in one instance) at best allow for ~20 stops of DR and brackets can be spaced by 3 stop or more, how much exposure bracketing is really beneficial?

Of course, as a Canon user I can only dream about 14 stops at ISO100.

-h

Re: How do you bracket your exposures?

PostPosted: Wed May 06, 2015 7:10 am
by Sebastian Nibisz
The gradients are less precise when the step is large. I do not recommend step greater than 2ev.

Re: How do you bracket your exposures?

PostPosted: Wed May 06, 2015 12:24 pm
by phil68
I mostly use 1eV Step, sometimes, in high contrast situations, I use 0.7eV
The results of the HDR picture strongly depends of these steps.

Regards,

Phil

Re: How do you bracket your exposures?

PostPosted: Thu May 07, 2015 9:54 am
by Joghi
For night shots, especially urban sceneries with bright street lights, i get the best results with bracketing: -6, -4, -2, 0, +2 EV. Depending on the scenery sometimes i donĀ“t need the +2.

Image

greez Joghi

Re: How do you bracket your exposures?

PostPosted: Mon May 18, 2015 9:54 am
by hjulenissen
Sebastian Nibisz wrote:The gradients are less precise when the step is large. I do not recommend step greater than 2ev.

Why is that? Guillermo seems to report excellent results with brackets spaced as wide as 4 stops (for an 8-stop DR camera), claiming that as long as the SNR is sufficient in all brightness regions, the end-result is good.

Fine-tuning of the exposure correction in the synthesis seems to be his idea of the key (not trusting camera exposure time/raw-file offset 100%).

http://www.guillermoluijk.com/article/v ... dex_en.htm

Re: How do you bracket your exposures?

PostPosted: Mon May 18, 2015 9:45 pm
by Sebastian Nibisz
hjulenissen wrote:Why is that? Guillermo seems to report excellent results with brackets spaced as wide as 4 stops (for an 8-stop DR camera), claiming that as long as the SNR is sufficient in all brightness regions, the end-result is good.
Fine-tuning of the exposure correction in the synthesis seems to be his idea of the key (not trusting camera exposure time/raw-file offset 100%).
http://www.guillermoluijk.com/article/v ... dex_en.htm

Let us assume that the camera sensor can register 32 levels of brightness (5ev range). Gradient will be mapped with poor precision in darker areas:
0ev, 16 levels (16-31)
-1ev, 8 levels (8-15)
-2ev, 4 levels (4-7)
-3ev, 2 levels (2-3)
-4ev, 1 level (1)

We can make two images with exposure adjustments (+1ev, +2ev). The gradient precision will look like this:
-1ev, 16 levels (16-31)
-2ev, 8 levels (8-15)
-3ev, 4 levels (4-7)
-4ev, 2 levels (2-3)
-5ev, 1 level (1)

-2ev, 16 levels (16-31)
-3ev, 8 levels (8-15)
-4ev, 4 levels (4-7)
-5ev, 2 levels (2-3)
-6ev, 1 level (1)

If all the images will be merged we obtain:
0ev, 16 levels (48-63)
-1ev, 16 levels (32-47)
-2ev, 16 levels (16-31)
-3ev, 8 levels (8-15)
-4ev, 4 levels (4-7)
-5ev, 2 levels (2-3)
-6ev, 1 level (1)

If we ignore the second image (+1ev):
0ev 16 levels (40-55)
-1ev 8 levels (32-39)
-2ev, 16 levels (16-31)
-3ev, 8 levels (8-15)
-4ev, 4 levels (4-7)
-5ev, 2 levels (2-3)
-6ev, 1 level (1)

Re: How do you bracket your exposures?

PostPosted: Sat May 23, 2015 9:15 am
by Bart_van_der_Wolf
Sebastian Nibisz wrote:The gradients are less precise when the step is large. I do not recommend step greater than 2ev.


Yes, I can confirm that.

As a test for the fusion method that SNS-HDR uses, I once made a test series of shots of a very high contrast scene (a automobile in a carport with half of the car in the full sun with reflections on the paint-coating and half in the shadow including the rear wheels/wheel arch). The series was made with 1/3rd stops intervals, and ranged from least exposed to avoid highlight clipping (I checked the Raw data), to most exposed to have a very high exposure level of the deepest shadows in the wheel arch.

I then let SNS-HDR fuse the results based on 16-bit TIFFs (straight Raw conversions, only whitebalanced, and with a linear tonecurve). I used series of 1/3rd stop intervals, 2/3rd stop intervals, etc. until I started actually seeing transitions between bracketed exposures, like sudden changes in noise structure in the car's smooth paint. It was at the 2 and 2 1/3rd stop intervals that I clearly noticed differences in the medium exposed scene regions.

In normal cases I now bracket with 1 or 1.33 stop intervals.

The exposure fusion that SNS-HDR seems to use is somewhat different from the 'Zero noise' approach by Guillermo Luijk. Guillermo basically uses the shadow bracketed exposure, and uses a relatively sharp cut-off mask where the Raw data clips, and then he uses the -4EV exposed shot with an inverted mask after very accurate exposure matching, to fill in the clipped portion of the shadow bracket. That can work, because it uses the highest exposed parts of both images (which then have the best Signal to Noise levels, mostly photon shot noise), but it requires very accurate Raw exposure matching (with floating point precision) to avoid sudden posterization at the transition, and depends on a predictable veiling glare performance of the lens.

SNS-HDR does a tremendous job, and by using TIFFS as input you get the benefit of the best Raw conversion tools for one's camera (some cameras require special conversion software, with darkframe corrections and flatfielding or colorcast calibration), specific camera ICC color profiles, the possibility to use lens corrections and Chromatic Aberration correction, and even the possibility to use different White-balance for the input images (e.g. warmer shadow exposure conversions and normal sun lit exposure conversions, or mixes with artificial interior lighting and sunlit daylight exteriors).

I'm looking forward to Version 2 of SNS-HDR which promises to be even better at several aspects.

Cheers,
Bart