The Never Ending Holy Quest for Dynamic Range

[ChristopherBarrett]

I hate blending HDR in post.  I do.  I really, really hate it.  Ok?  Ok.  So...  I've spent the better part of the morning on a quest for as much dynamic range as I can get out of a single capture.  Here are my results...

Image 01  P65+ with pretty straight settings in C1 Pro (The Control Sample)


Image 02  P65+ with Curve set to Linear Response, HDR Sliders and Clarity at 20


Image 03  P65+ processed in ACR.  The color science is ass but WAY more hilight info!


Image 04  RED Epic, pretty likely my favorite.  Extended range with a natural feel.


When you look at the image processed in ACR, it's obvious that there is a lot more hilight info in the Raw than C1Pro will give me.  What up C1 Pro?  Why you holdin' out?

Doug?

[design_freak]

My favorite chair :-)

[Slobodan Blagojevic]

Funny, but i find the first one the most pleasing and natural appearing, blown highlights or not. All others look muddier, with mid-tones underexposed.

[Kirk Gittings]

Great test CB. I am always looking at this DR issue to see if I can improve my produce. What was the braketing for the HDR? And what program?

I agree SB, though a touch more detail in the curtains would be nice. The others' midtones could be perked up pretty simply IME. What would I do? FWIW-layer and blend the top version with the curtains form one a bit brighter than the HDRs.

[ChristopherBarrett]

Actually there is no bracketing.  I'm trying to squeeze as much out of a single exposure as I can to avoid the whole HDR blending.  So the point really isn't "which of these is initially more pleasing" but "which process is the most successful approach to expanding the dynamic range of a single capture."  Any of the examples with greater curtain detail could be adjusted for more mid-tone contrast once the hilights are established.

I constantly run into situations where the client wants an interior to feel bright and airy while maintaining the view.  The blown out Crate & Barrel look, while a nice esthetic, doesn't resolve the client demands.

So.... rather than sticking to my current approach of hand blending a window view into a normal exposure, I want to find the best way to capture all the DR I need in one exposure which I can easily add mid-tone contrast to later.

BTW... Tony Kuyper's Luminosity Masks are essential (in my opinion) for this sort of fine tuned image manipulation.

Yada yada ya

[EricWHiss]

CB,
Have you tried a multishot back?  For that kind of work you show here, the MS is the way to go as you get 1-2 stops more useable DR. 

[haefnerphoto]

Chris, Is the software for the Red Epic capable of processing P65 files?  It did look the best to me also.  Since you're not shooting anything that's moving you should try Photomatix, blending exposures using the Fusion blend mode.  I find it works really well and is a great base for your final imaging.  Jim

[ChristopherBarrett]

Thanks for your thoughts...  I often place people in alternate versions of our shots, so I'm not sure about using a multishot back...  I've always been intrigued by their image quality and sharpness, tho. 

The Red files can be processed in their software or a number of other suites, RedCineX doesn't work with other file types.

My typical process is to use Photomatix with Exposure Fusion... I then take the resulting file and usually layer it into a normal contrast file, hand blending the trouble spots (Windows and light fixtures).  This method is pretty decent, but can be time consuming and it doesn't usually leave me fully satisfied.

[haefnerphoto]

Thanks for your thoughts...  I often place people in alternate versions of our shots, so I'm not sure about using a multishot back...  I've always been intrigued by their image quality and sharpness, tho. 

The Red files can be processed in their software or a number of other suites, RedCineX doesn't work with other file types.

My typical process is to use Photomatix with Exposure Fusion... I then take the resulting file and usually layer it into a normal contrast file, hand blending the trouble spots (Windows and light fixtures).  This method is pretty decent, but can be time consuming and it doesn't usually leave me fully satisfied.

Chris, I don't know of a better workflow.  We're able to produce images in lighting situations that we would have never tried in the past with better than acceptable results.  The downside is that the time spent in post is at least as long as the time spent shooting (at least for me) and unfortunately, not billed at the same rate.  I don't think great light will ever be topped but with the post processing available to us it's hard to take a bad shot (at least from a lighting perspective).  I often compare my fused images to the bracket they came from and never do the individual captures exhibit the same dynamic range and if I try to adjust an individual capture in C1 it helps but still isn't as good.  Incidentally, this is a solution that is more often used in architectural photography than other specialties.  I rarely utilize this workflow with my automotive projects because I'm more often than not lighting the subject.  Attached are two images that I'm happy with that didn't require a lot of post manipulation, the light was right and I was there.  Jim

[Kirk Gittings]

Nice Jim.

CB, I find it necessary to do an Enfuse/Lightroom blend on a room EVEN if I will be manually inserting an outside view in a window. Why? because the LF/Enfused image will give me better, believable, window frame without all the flaring from overexposure. You?

[ErikKaffehr]

Hi,

That is tone mapping, the art (not science) of inserting a large square object into a smaller circular hole ;-)

I prefer the last image, processed by the RED software, BTW.

My guess is the problem gets even worse, if we want to print, as a screen can hold a contrast range of perhaps 1:500 but paper is limited to around 1:150.

Best regards
Erik

Funny, but i find the first one the most pleasing and natural appearing, blown highlights or not. All others look muddier, with mid-tones underexposed.

[David Grover / Capture One]

Thanks for your thoughts...  I often place people in alternate versions of our shots, so I'm not sure about using a multishot back...  I've always been intrigued by their image quality and sharpness, tho. 

The Red files can be processed in their software or a number of other suites, RedCineX doesn't work with other file types.

My typical process is to use Photomatix with Exposure Fusion... I then take the resulting file and usually layer it into a normal contrast file, hand blending the trouble spots (Windows and light fixtures).  This method is pretty decent, but can be time consuming and it doesn't usually leave me fully satisfied.

Hi Chris,

As a 4-shot and a 1-shot are the same size it is easy to shoot the interior as a multi, and as long as your camera angle doesn't change, shoot several singles of your model in place, pick the best one and drop them in quicly on a layer mask.

Best of both worlds. 

David

[eronald]

Funny, but i find the first one the most pleasing and natural appearing, blown highlights or not. All others look muddier, with mid-tones underexposed.

I agree. The first one is the best looking image, the last one might recover with a bit of processing. I've heard RED starts out very flat, to allow for color grading.

Edmund

[haefnerphoto]

Nice Jim.

CB, I find it necessary to do an Enfuse/Lightroom blend on a room EVEN if I will be manually inserting an outside view in a window. Why? because the LF/Enfused image will give me better, believable, window frame without all the flaring from overexposure. You?

Kirk, How do the Enfuse/Lightroom blends compare to Photomatix?  Does it do a better job with handling the window detail flare?  My workaround for handling that is to shoot the window area with strobe to balance the frame detail with the outside exposure which adds a fair amount of time to the shot.  Jim

[ondebanks]

As to which version works best? Here's how I would see it: Which rendering shows the house that you would rather live in? I would definitely prefer to live in the bright, airy, first one. The 4th one isn't bad either. The 2nd & 3rd ones say to me: even though there's daylight coming in from both sides, this is such a gloomy room that the lamps have to be turned on to brighten it up!

I guess people are so irrational: they do not choose their living quarters on the basis of how well compressed the dynamic range is. 

Ray

[yaya]

Chris I have to agree with Edmund and Ray, the first image looks like it was shot on a nice sunny day. The HDR ones look like they were shot in London on a cloudy winter day. It is good though to see the LR's highlight recovery has been improved in recent versions

I don't think a multi shot back will be so practical for you as it'll require you to remount your lenses into electronic shutters and to drag more kit: cables, batteries, power supply etc. The file will still be smaller than what you're getting now and you won't get the same AOV what you already have

All IMO of course

yair

[ejmartin]

What up C1 Pro?  Why you holdin' out?

It's not a question of 'holding out'.  One or more channels are blown out, and the software has to guess what is a reasonable value for the blown channel based on the values of the non-blown channels.  Color accuracy is the first thing sacrificed, since an inaccuracy in the guess yields inaccuracy in the color.  Luminance information is somewhat easier -- just copy the structure of the unblown channels into the luminance channel.  Anyway, different software programs have different methods of guessing, with varying results.

[torger]

Digital is not too good at highlights. There's no dynamic compression like in film, if there's too much light it just clips. And for RGB sensors the channels don't clip at the same time, so you lose one channel at a time around the highlight.

If you really want to maximize the dynamic range in a shot you can use colored filters on the lens to balance the sensor's channels for a particular light setting (so they clip more or less simultaneously), and apply correction for that in post. Few software programs allow to watch the actual RAW histograms (and RAW clipping) though so it is hard to see exactly how a picture is exposed on the sensor.

Some RAW processors have advanced guessing algorithms to "restore" highlights (actually guessing the missing information). In difficult conditions when I've overexposed some I use RawTherapee which allows to switch between different algorithms, since it depends on the photo which algorithm that works the best.

Fortunately it is very common that highlights are actually white, and then highlight reconstruction algorithms can perform really well. But as said, it is in part guessed information so authenticity will suffer. Standard software with standard processing settings will play it safe and not try to extrapolate the highlights. When highlight reconstruction algorithms that tries to extract the most out of the given information gets the guessing wrong it can look really bad.

[fotometria gr]

Actually there is no bracketing.  I'm trying to squeeze as much out of a single exposure as I can to avoid the whole HDR blending.  So the point really isn't "which of these is initially more pleasing" but "which process is the most successful approach to expanding the dynamic range of a single capture."  Any of the examples with greater curtain detail could be adjusted for more mid-tone contrast once the hilights are established.

I constantly run into situations where the client wants an interior to feel bright and airy while maintaining the view.  The blown out Crate & Barrel look, while a nice esthetic, doesn't resolve the client demands.

So.... rather than sticking to my current approach of hand blending a window view into a normal exposure, I want to find the best way to capture all the DR I need in one exposure which I can easily add mid-tone contrast to later.

BTW... Tony Kuyper's Luminosity Masks are essential (in my opinion) for this sort of fine tuned image manipulation.

Yada yada ya

Have you tried film for this particular shot? I do use a multishot back and it wouldn't work in this, (curtains always move, even if its negligible to ...some). I'm also a "DR Freak" whatever this means, I never use HDR because its unnatural in my view and I don't like the RED image because of the unnatural looks of the books, I also not consider DR as low lowlight extension, but only as highlight. IMO, the image I would prefer would be the first one, with the RED's curtain and stand beside it, I think that my Contax 645 with old Fuji 160s would achieve that, I would also try my 528c (in single shot), but only for the purpose of trying it, I wouldn't trust digital to solve this particular problem, particularly because I feel that your view on the lighting coming from the window (and hence the curtain appearance) is very important for the scene. Regards, Theodoros. www.fotometria.gr

[fotometria gr]

Digital is not too good at highlights. There's no dynamic compression like in film, if there's too much light it just clips. And for RGB sensors the channels don't clip at the same time, so you lose one channel at a time around the highlight.

If you really want to maximize the dynamic range in a shot you can use colored filters on the lens to balance the sensor's channels for a particular light setting (so they clip more or less simultaneously), and apply correction for that in post. Few software programs allow to watch the actual RAW histograms (and RAW clipping) though so it is hard to see exactly how a picture is exposed on the sensor.

Some RAW processors have advanced guessing algorithms to "restore" highlights (actually guessing the missing information). In difficult conditions when I've overexposed some I use RawTherapee which allows to switch between different algorithms, since it depends on the photo which algorithm that works the best.

Fortunately it is very common that highlights are actually white, and then highlight reconstruction algorithms can perform really well. But as said, it is in part guessed information so authenticity will suffer. Standard software with standard processing settings will play it safe and not try to extrapolate the highlights. When highlight reconstruction algorithms that tries to extract the most out of the given information gets the guessing wrong it can look really bad.

+1. I already suggested film at the same time you was quoting as well, I feel however that digital will overcome its problems (sometime in the future...)  and then we maybe all seek the resolution we want  Regards, Theodoros. www.fotometria.gr
P.S. The question is what we are gonna do with the values worth of our super duper cameras at THAT DAY.