High Dynamic Range Imaging

[Mike Peregrine]

I just thought of something else too ... lol ... I'm thinking that the more exposures a person could take, the better. Like a range of 7-10 is better than a range of 3. If you have a set of three, with an exposure at a stop above and a stop below, then the software may tend to want to pull from the extremes of the two, which would cause the image to whack out a bit more.

If you are running 5-7 different exposures, though - the software will tend to blend from a wider range and the photo should come out looking more balanced in the end. Just an idea - - will be tough to do in a storm environment, and a person will have to be quick. But I would think with a typical landscape (with as little movement as possible), a person should be able to get at least 5 exposures within a split second of each other. Add people, cars, or lots of motion in the cloud base (i.e., a TORNADO), and this probably wouldn't work without losing focus.

 

[Ryan McGinnis]

Just a couple quick things -

First, this thread rocks. Like, a lot. Ryan - thanks for going through the effort of Beta testing this method and posting your results. You can bet your buns that there's more than one photog interested in trying this out.[/b]

Thanks, Mike! I thank Darren for pointing this whole HDR thing out to me to begin with. I really do feel ignorant for not knowing about it, given as much as I use Photoshop.

One question for you, though - how many exposures did you take to achieve that shot above? - and if I'm understanding this process right, and a person brackets 3 or 4 exposures that the software can then "merge" to draw the best levels from each, how do you avoid blur or camera shake? - In your images, your subjects are crisp. How did this work if you are dealing with a number of exposures? I've been trying to read up on this, but still haven't figured that point out. [/b]

In this particular shot, I took seven images, each seperated by one stop. My camera's aperature dial moves in 1/3rd stop increments, so I just tripoded the camera, attached a cable release, took a shot, clicked the dial three clicks, took another, lather, rinse, repeat. Altogether it took about seven seconds to do, according to the EXIF info. I've found that one can actually get very useable HDR images just by doing the standard 2 stop bracketing, which takes no time at all, since most Canon digital bodies have a setting that will do this for you simply by holding down the shutter release until three shots have fired.

The reason that nothing looks blurred here is basically because nothing much moved. If you look close at the foot of the guy hanging out of the SUV, you'll see that it is a bit blurred in the HDR shot, as he did slightly change position over the seven seconds it took to do the shots. Also, the clouds above look a little blurred. But using the 10mm side of the lens (effectively 16mm) means that something has to move quite a bit to move much in the frame. The storm was also really slow moving, which helped.

Here's some thumbnails of the seven shot RAW sequence that I fed into Photoshop. All should have EXIF info still in them. The timestamp is probably a bit off as I need to set my camera's time again, but everything else should be right.

I've been enjoying this thread from the sidelines as well. I'm not very good at digital processing, I only switched over to a digital body this last winter. This technique looks quite interesting - and I'm eager to try it out on a few images I can think of off the top of my head. Anyhow - my question is for Ryan regarding the image above: what lens are you using? The perspective distortion is pretty bad on the vehicles - so I'm guessing it's some sort of fisheye lens, which you are then processing to convert back to rectilinear. [/b]

For that shot I used the Canon 10-22 at f/8. The distortion is pretty bad, but it's actually not a fisheye, it's a rectilinear lens. It's just that at an effective 16mm, distortion is always going to come into play. From what I understand, the 10-22 controls distortion pretty well for a superwide.

I would agree the foreground is too bright - but I think that mainly is disturbing because of the shadowed treeline on the horizon - that gives the illusion that the foreground has some artificial lighting supplement. There also seems to be some color halos in the cloud edge near the upper left - which makes me wonder if this was a slow bracketed shot. [/b]

I actually burned down the treeline because I thought it looked fake when it was properly exposed! I was trying to create a deliniation between the sky and the ground. I think that Mike U. is right -- HDR toning really is an art, and I've got a lotta learnin' to do.

 

[Mike Hollingshead]

Just as an example of how far you can go with one single exposure in just photoshop. If you want to be able to take one shot and be able to open it to where it should be I think your 4th exposure down is the best one. The sky is just barely starting to blow out but it is still very much intact. Right there you are able to open the foreground past where you would even need to. You can also get contrast in the storm and the foreground while keeping the brighter sky in check.

I just think you can get a ton of tone and info from one exposure but it would have to be just right.

Here is the same 4th one down with the foreground darkened back down a bit from where I took it in the example...

 

[Ryan McGinnis]

That's true, one does have a large amount of latitude to do manipulation, especially if you shoot RAW. For the most part it works, but after a certain point you run into combing histos and banding. This becomes a much bigger deal when you make enlargements. If you're not going to tone map, then just straight doing a bracketed exposure and then composting the layers in Photoshop would be the way to go, as this would avoid running the risk of banding. Of course, this only works if you have time to set something like that up or can hold the camera rock steady -- which isn't always the case when you're near storms.

Below are a couple of examples. One is a 100% crop of the image that you picked, exported to a 16 bit TIFF and opened up with levels and curves to about the point you were indicating with the preview JPEG. Noise and artifacts are starting to show up, but I would say that this is pretty acceptable for a 100% crop. I doubt this would look bad in a print. The other is a 100% crop of the same area of the HDR image. You can see the shortcomings of HDR here: as the person moved, it messed up the HDR process. (There is also some bizarre color banding the comes in to play if whatever you shot moved in between bracketed shots.) It also emphasizes to a high degree any chromatic abberation that your lens is experiencing (I didn't do any CA correction in the RAW import, though I probably should have.) However, you can see that the artifacts are gone and the shadows fade smoothly away. For a completely static scene, HDR offers a lot, though you can definately create an excellent image without the HDR. Ultimately, with that many exposures, if I wasn't going to tone map the best thing to have done would have been to have taken the third shot down, opened it up a bit, and used it for the ground, then composted the sky from the fourth image down on top of that. Between the two, I'd have had all the dynamic range I'd really have needed.

 

[Darren Addy]

...has anyone thought of trying this with good old b&w yet? - This may really work well for black & white. You won't have the questions about unnatural color, but you'll still enjoy the wide latitudes. I've already ordered PS-CS2 and think I'll give it a whirl just to see what the results are like.
[/b]

Mike,
I said this to Ryan on our way down to Hays on Sunday. Tone mapping is essentially Ansel Adams "Zone System" for color. If that looks "unrealistic" to you, try making it into a black and white. Become the Ansel Adams of storm chasing!

The other thing that I will say about this is that HDR seems much more "honest" to me than some of the "over-processing" (as some might call it) being done with Photoshop alone. Instead of basing your improvements on some algorithm, with HDR you are using real details from the scene (albeit at bracketed exposures) and pulling the details out of each of them.

Kudos to Ryan for showing us where this might go — and he's been doing it now for [what?] a week? I fear that I have created a monster. :blink:

Darren Addy
Kearney, NE

 

[Oscar van der Velde]

I just think you can get a ton of tone and info from one exposure but it would have to be just right.

Here is the same 4th one down with the foreground darkened back down a bit from where I took it in the example...
[/b]

This is what I mean how the image is processed to look better and still looking realistic to my eyes/brain. Very good job on that, Mike! If one can get this out of a web jpeg, imagine what's possible with RAW. Good enough range to not have to go through the inconvenience of having to take multiple shots, while you can use the space on the card for something else

Oscar

 

[Ryan McGinnis]

This is what I mean how the image is processed to look better and still looking realistic to my eyes/brain. Very good job on that, Mike! If one can get this out of a web jpeg, imagine what's possible with RAW.
[/b]

Hi Oscar -- no need to imagine, look up! I used RAW to create the first 100% crop image above. That's what's possible with RAW with that shot.

 

[Darren Addy]

Those of us not yet Photoshop Gurus (or maybe just interested in extending your abilities with it) may find Photoshop for Photographers useful. It's a pretty complete introduction to Everything You Ever Wanted to Know and includes a section on HDR (under Basics).

Darren Addy
Kearney, NE

 

[Cstok]

This may be a dumb question, but if you shoot in RAW, couldn't you just shoot one picture, and then use Adobe Camera Raw, or some other app and save 3 or 4 pix with different settings and then combine? Or do you have to actually shoot multiple picutres.

Great topic by the way!

 

[Mike Peregrine]

This may be a dumb question, but if you shoot in RAW, couldn't you just shoot one picture, and then use Adobe Camera Raw, or some other app and save 3 or 4 pix with different settings and then combine? Or do you have to actually shoot multiple picutres.[/b]

I had this question at first too - and I think from the sounds of the things I've read in the other forums on the subject that this will become a reality - and working from a single RAW image (which would be the best choice) might be able to work with this method - it doesn't sound like it's quite there yet. The only things I'm wondering about involve the fact that when you use a camera's manual settings to actually take separate photos, you are picking up detail in each exposure (especially when it comes to things like toning and color) that would become lost in a computer's attempt to render the same effect (by simply adjusting brightness/contrast or whatever). It still looks like separate exposures are the way to go, for the time being anyway, to get the full benefit. You could REALLY get a shot to look whacked out if you started photoshopping a RAW image and use a blend of different exposures PLUS altering the color of each shot. I wonder if the computer would even know what to do with it. I bet a person could end up with some pretty weird effects.

 

[Mike Hollingshead]

This may be a dumb question, but if you shoot in RAW, couldn't you just shoot one picture, and then use Adobe Camera Raw, or some other app and save 3 or 4 pix with different settings and then combine? Or do you have to actually shoot multiple picutres.

Great topic by the way!
[/b]

We already kind of mentioned the problems with doing this above. Read Glen and my post about the noise problem. I think you can close something down with the exposure compensation but I have yet to see any real benifit from using it to open something up. It seems to do no more than simply using photoshop to open up the image. Perhaps you could do it with a very over-exposed image, but I've not tried to see how much you can fix blow outs by stopping down after the fact in conversion. I've see you can fix some degree of over-exposure with doing that but not sure if you can drop 4 stops in conversion and have things be ok....I'd kind of think you should be able to. So again, as far as changing exposure in conversion after the shot I don't think it does a whole lot extra as opposed to doing it with photoshop. You'll gain noise and be lacking info having to open something up either way...at least it seems so to me.

 

[Oscar van der Velde]

If the technique with just one RAW image would get you grainy shadows, it should be possible to just use two exposures: one exposed so that you just get all of the highlights, and one exposure maybe 1-2 stops brighter for the shadows. I don't see how one can get more info if the exposures overlap in the brightness range so much, as in the above example series from Ryan. The shots which put the maximum highlights farther into the left of the histogram contain less and redundant info for sure, since you effectively only raise the black point in the histogram, shadows becoming more and more true black.
Two or three shots are on most DSLRs taken quickly and easily with the auto-bracketing function and with the camera set in the correct drive mode (multiple shots in a row). Personally I would do it in very contrasty situations with more than 3 stops difference between foreground and the highlights you want to render with detail. Needless to say, using a tripod!

Oscar

 

[Ryan McGinnis]

Thought I'd point out another HDR storm photo (not mine):

Storm over San Francisco

The Photomatrix halos drive me nuts, but wow, that's a spectacular shot.

 

[Ryan McGinnis]

BTW, I downloaded the Photomatix trial and fed it the HDR base shots that I used to cread the HDR image earlier. Here's what the software spit out when it was set to it's thermonuclear "kick this image's ass!" setting:

Looks cool, but very unrealistic and with with tons of halos everwhere. And I feel dirty for doing this to a photo -- it looks nothing at all like what I saw when I was there.

 

[Mike Umscheid]

BTW, I downloaded the Photomatix trial and fed it the HDR base shots that I used to cread the HDR image earlier. Here's what the software spit out when it was set to it's thermonuclear "kick this image's ass!" setting:

Looks cool, but very unrealistic and with with tons of halos everwhere. And I feel dirty for doing this to a photo -- it looks nothing at all like what I saw when I was there.
[/b]

LOL Ryan... Thanks for testing this out... if nothing else this makes me feel a little better doing all my blending by hand in Paint Shop Pro/Photoshop.... Ahhhhh algorithms.. The only kick-butt algorithms are things like digital noise reduction and unsharp mask