Input to the HDR software - created exposures

[BarryTX]

I'm looking for opinions based on actual experience. The question is, if you did not bracket a shot but did take it in RAW, is it better to create additional exposures in your RAW processing software - a +2 EV and a -2 EV for example - so you can feed your HDR software multiple exposures? I use Photomatix and have played with creation of a 'pseudo-HDR' image both from the method above and from just the single photo. I've also done it with RAW directly in Photomatix as well as with jpg, which they claim is just fine.

What I have seen so far makes me believe that RAW or tiff as input in Photomatix is better than jpg, better being defined as the final image looking more realistic and not over-saturated (which I usually want to avoid). I also feel like the 'created' extra exposures results in a better (again, more pleasing to me) final tonemapped image than the single file alone.

Any thoughts here? What are your experiences?

[carregwen]

I'm looking for opinions based on actual experience. The question is, if you did not bracket a shot but did take it in RAW, is it better to create additional exposures in your RAW processing software - a +2 EV and a -2 EV for example - so you can feed your HDR software multiple exposures? I use Photomatix and have played with creation of a 'pseudo-HDR' image both from the method above and from just the single photo. I've also done it with RAW directly in Photomatix as well as with jpg, which they claim is just fine.

I've done quite a few. I always use RAW. You can create +- exposures from one normal, but I find it works slightly better with bracketing. But there isn't all that much in it. If you do it from one you obviously don't need to do the alignment stuff in the processor, which is a good thing. I suppose a lot depends on the subject - if it's moving it's impossible to bracket and merge.

This one was done from three separate exposures. http://www.flickr.com/photos/carregw...7623631158436/

I'm not sure many people know this, but on most cameras if you set up bracketing, then set auto-timer, it will open the mirror, take the three shots without closing the mirror, then close it when finished. It's very quick on my Canon 50D and 5D, and avoids camera shake - but obviously it needs to be on a tripod. I use Dynamic Photo HDR, which I find is very good.

[Dave Humphries]

I'd be amazed if 'doing HDR' with a single jpg, which is only 8 bit input, would achieve anything, because the output is also 8 bit.
At least, nothing you couldn't do quicker with just a S curve or similar adjustment to re-tone map things slightly.

I can't better Rob's thoughts on single RAW, so I'll pipe down now

[Colin Southern]

Hi Barry,

In terms of increased dynamic range, altering a single RAW is an "old wives tale" I'm afraid, as the altered shots can only contain LESS information, not more (you can't make a silk purse from a sow's ear). However - doing it that way may well trick the likes of Photomatix.

I know that Photomatix can produce realistic looking HDR images, but by default, it seems to produce an over-saturated over-sharpened tone-mapped image that - unfortunately - many are confusing with "a HDR look" (HDR doesn't have a "look" - it's simply a set of strategies for CAPTURING a scene when the dynamic range of the scene is too big to capture in a normal single exposure).

We've had a "robust" discussion on this in the past that you might find helpful, and I also wrote a mini-article on tone mapping that seems to have helped a lot of people on the topic as well.

[carregwen]

In terms of increased dynamic range, altering a single RAW is an "old wives tale" I'm afraid, as the altered shots can only contain LESS information, not more (you can't make a silk purse from a sow's ear). However - doing it that way may well trick the likes of Photomatix.

I know that Photomatix can produce realistic looking HDR images, but by default, it seems to produce an over-saturated over-sharpened tone-mapped image that - unfortunately - many are confusing with "a HDR look" (HDR doesn't have a "look" - it's simply a set of strategies for CAPTURING a scene when the dynamic range of the scene is too big to capture in a normal single exposure).

You are quite right there, Colin. A single RAW can never match the actual HDR range that multiple exposed shots can. But, the tone-mapping can sometimes (as long as you don't overdo it) create the HDR effect (but it's not really HDR). For some reason many people seem to like that look. Personally, I find it often looks over-cooked. The problem with multiple shots is it can't be used for moving subjects, although the shooting method I gave above using auto timer does speed up the elapsed time between the three shots.

I'm out taking shots this afternoon. I'll give it a go with one shot and three and post the results.

[arith]

It's raining here rob. Since getting ACR I find it hard to justify HDR at all; previously I could tone map a noise cleaner image than just raising shadows ' fill light in ACR'.

But still, some of my images were over 13ev span but not many. Incidentally the HDR software I use finds it difficult to compress images with ev span greater than 13 but I only really need it for natural looking photo's and the commercial paid for ones are really difficult to get natural. Picturenaut is really easy to do natural and I like easy.

[Colin Southern]

You are quite right there, Colin. A single RAW can never match the actual HDR range that multiple exposed shots can. But, the tone-mapping can sometimes (as long as you don't overdo it) create the HDR effect

I think you mean the "over-processed, photomatixed effect"

For some reason many people seem to like that look.

To be honest, I don't mind it myself (in relatively small doses) -- I've just dedicated the rest of my natural life to helping people disassociate "that look" (or "the HDR look") with HDR.

The problem with multiple shots is it can't be used for moving subjects

It depends on the technique used, and where the movement is; if you're using a technique of overlaying 2 or more images and then creating a transition mask then movement is generally fine so long as it's not in the transition zone (I say "generally" because you still have to watch reflections not matching etc).

[Colin Southern]

It's raining here rob. Since getting ACR I find it hard to justify HDR at all; previously I could tone map a noise cleaner image than just raising shadows ' fill light in ACR'.

But still, some of my images were over 13ev span but not many. Incidentally the HDR software I use finds it difficult to compress images with ev span greater than 13 but I only really need it for natural looking photo's and the commercial paid for ones are really difficult to get natural. Picturenaut is really easy to do natural and I like easy.

Hi Steve,

A 13EV delta is probably about 1 stop over the max for most modern DSLRs - and even then you'd have pretty noisy shadows even if you did nail the exposure. Personally though, I like to use a 3-Stop GND filter in those situations - whist means that that camera is only seeing a 10 EV span which you can get away with. If you can combine multiple exposures then you can get the noise significantly better again.

eg this shot ... 3x 30-sec exposures for the rocks - single exposure for the sky (all same exposure) - sky tweaked differently in ACR (simply to make the transition easier - no increase in DR as we all know) - and all shot with 3-Stop hard-edge GND.

[carregwen]

Very good shot, Colin. Although for some reason I'm getting an optical illusion (no, not squirrels) I keep thinking the rocks are underwater. I know that's silly as you would never put your expensive black box underwater - even though it's waterproof.

I shouldn't have said that - you'll be seeing it now!

[Colin Southern]

Very good shot, Colin. Although for some reason I'm getting an optical illusion (no, not squirrels) I keep thinking the rocks are underwater. I know that's silly as you would never put your expensive black box underwater - even though it's waterproof.

I shouldn't have said that - you'll be seeing it now!

Well ... the rocks WERE under water (sort of / kinda). Not wishing to burst any bubbles (*** spoiler alert ***) there's really no mist in the shot at all - it's just the effect you get with water breaking over rocks with a long exposure

[arith]

I didn't spot that; cheers Colin

[BarryTX]

Hi Barry,

In terms of increased dynamic range, altering a single RAW is an "old wives tale" I'm afraid, as the altered shots can only contain LESS information, not more (you can't make a silk purse from a sow's ear). However - doing it that way may well trick the likes of Photomatix.

>

This is what I am wondering - it seems to me that the extra, created exposure files do help the output in Photomatix. If this is true, and not just my imagination, the software appears to be getting more data out of the 3 files than out of 1. It would make sense that this would be a software artifact if indeed true.

I assume the RAW file created by my Canon G11 has as much info in it as the camera could record, so I did not expect the created exposures to add anything (unlike actual bracketed exposures) but the more I'm playing with Photomatix the more I'm becoming convinced that it does help with this program.

Thanks to all for the responses, and great shot Colin!

[pwnage101]

With just one raw image I think you can import it directly into photomatix as an individual image (If I can recall correctly). Otherwise, you can use photoshop's "shadows/highlights" feature. If you do this, make sure the you export from ACR with blacks and whites nearly clipped, and as a 16 bit image.

[Colin Southern]

>

This is what I am wondering - it seems to me that the extra, created exposure files do help the output in Photomatix. If this is true, and not just my imagination, the software appears to be getting more data out of the 3 files than out of 1. It would make sense that this would be a software artifact if indeed true.

Hi Barry,

It's not getting "more data" per sec because you can't extract information that isn't there to start with - but - it can certainly alter the result, and quite possibly "for the better" if "better" is defined as "more pleasing to the creator". "Artifact" is an interesting choice of word though; normally I'd define "artifact" along the lines of "something that appears in the result that wasn't in the source" ... which (no disrespect to Photomatix creators) pretty much sums up what Photomatix tends to produce by default ... ie something that's pleasing in the eye of many beholders, but bearing little in common with photo-realism. In this regards it appears to be 5% HDR (in that it's combining a high dynamic range of source images), but 95% "Ultra-tone Mapping" (the process by which it produces it's "characteristic look").

[carregwen]

>

This is what I am wondering - it seems to me that the extra, created exposure files do help the output in Photomatix. If this is true, and not just my imagination, the software appears to be getting more data out of the 3 files than out of 1. It would make sense that this would be a software artifact if indeed true.

OK. This is a shot I did yesterday, and this is a snapshot of it straight out of RAW. I think it could win a prize for the most boring shot ever. I chose the scene as it had a lot of light and shade. The sky is burnt out. I did three exposures in-camera by bracketing. This was the middle (normal exposure) and so represents a shot you would normally take.



This is the same shot processed in ACR and PS/CS4. I tried to recover the sky, but couldn't.



And this is the 3 RAW files put through Dynamic Photo HDR to create a 16bit PSD file. Notice the sky, which looks quite natural even though it has recovered a lot of detail data. And notice the lack of what Colin refers to as the 'characteristic look'. It is not overcooked in my opinion. But others may disagree. I tried three exposures from one raw but the sky was just too burnt out and it looks awful.

The conclusion? Use bracketing and process three in-camera exposures!!!

[Colin Southern]

Hi Rob,

Photorealistic HDR - congratulations ... not often we see that

This was the middle (normal exposure) and so represents a shot you would normally take.

Possibly - but only if they're breaking the digital "expose for the highlights, develop for the shadows" rule.

Out of interest, what does the shot with the least exposure - but a dollop of fill light - look light?

[carregwen]

Out of interest, what does the shot with the least exposure - but a dollop of fill light - look light?

Not bad on first glance, although it's not as good as bracketing. But... the noise close-up is pretty bad compared to the noise in the bracketed version which is very clean. I think this shows that with bracketing and using the right options in HDR S/W you can get a realistic looking shot with an improved range.

[Colin Southern]

Not bad on first glance, although it's not as good as bracketing. But... the noise close-up is pretty bad compared to the noise in the bracketed version which is very clean. I think this shows that with bracketing and using the right options in HDR S/W you can get a realistic looking shot with an improved range.

Couldn't agree more. To be honest, that's the enigna I find with Photomatix; if you look at some of the examples in the websites of their members there are many many photorealistic examples -- and good ones too - and yet 99 times out of 100 whenever I see the typical "over-processed" look, it's also produced with Photomatix. All I can put it down to is people either not selecting the correct options, or not processing the image correctly (or at all) after Photomatix has done it's thing.

[Dave Humphries]

Hi Colin,

Can you clarify something for me (us) please?
I'm about to reveal my ignorance now, so bear with me while I stand in front of something red so you can't see me blush.

I've perused the Real World ACR book, but haven't found this definitively stated anywhere, perhaps because it's so blindingly obvious.

A single RAW file is typically 12 bits, or 14 if you're lucky, and has a linear gamma (1).
A jpg is 8 bits and has gamma of 2.2.
I don't think, having done a bit of research this morning, that gamma is at all relevant to what comes next, so I'll continue ignoring that.

To my question:
In an ACR RAW conversion to 8-bit, or similar conversion done later in workflow (in Photoshop or Elements), are the 12/14 bit pixel values (tone) mapped* to the 8 bits output?
* So 111111111111 (4095) becomes 11111111 (or 255) AND 000000000000 become 00000000 and everything in between is also linearly converted; so 2048, and 2047 and 2049, all become 128, etc.

Or does it take say, the central 8 bits, and just output those? (clipping 2 bits off both top and bottom)

If the mapping occurs, as I believe it does, then that makes the argument against HDR from a single RAW file more understandable to me, because all the image's pixel levels were in the 8-bit file anyway, albeit compressed/quantized together.

Producing 2 or 3 jpg (8 bit) output files from a single RAW file (with exposure biassed by say + 2 stops and blending them manually, or in a HDR program, may allow a more 'picture content aware'/selective area tone mapping to be done, but the gains are likely to be small - right? (and non-existant if in 16 bit mode?)

I know I should (and do) 'Open' from ACR to Elements in 16 bit mode (by changing that setting bottom centre of the ACR screen away from the default 8 bits), then continue working in 16 bit for as long as possible. Unfortunately, if you're working in Elements, there soon comes a time when you have to convert to 8 bit, for example, to use the clone brush. For CS3/4 users, this probably only occurs when you have finished PP and just want to Save As a jpg for display here at CiC. { where else? }

OK, I'm reddy for the answer,

[carregwen]

I know I should (and do) 'Open' from ACR to Elements in 16 bit mode (by changing that setting bottom centre of the ACR screen away from the default 8 bits), then continue working in 16 bit for as long as possible. Unfortunately, if you're working in Elements, there soon comes a time when you have to convert to 8 bit, for example, to use the clone brush. For CS3/4 users, this probably only occurs when you have finished PP and just want to Save As a jpg for display here at CiC. { where else? }

Don't go red yet. I'm always wondering this. What I did with the above test shot (and what I normally do) is load the 14bit RAWs direct into HDR program. Then merge them, then tone-map, then save as 16bit TIFF. Then I do as much as I can in 16bit before converting a copy to 8bit. I always thought that doing it this way you still get the spread of tonal range.