Is HDR imaging possible cheap?

K

kidgoldenarm

Member
Hey all,

This past weekend, I found a gallery to some amazing photos.

http://englishrussia.com/?p=532#more-532

Now these images were just gorgeous, I thought. But how in the hell did someone take these shots? I was like, there's no lens/film combo in the world that could produce these shots! I figured something must've been done in post to create these photos (a la photoshop or some other app).

Upon looking through the comments for the gallery, I found that a lot of people had deduced that the images were created using HDR (High Dynamic Range) Imaging.

I started to a little digging on the web, and found a little info, so I am anything BUT and expert, but this is the gist of what I managed to get from what I read on the topic.

Here's my stab at explaining the concept, and please anyone who know more, please jump in and take over.

As everyone who's ever taken photography 101 knows there are certain limitations that you run into when trying to get certain exposures. The three biggest obstacles that a good photographer learns to overcome/manipulate/wield are the F Stop (the size of the apeture opening), shutter speed, and ISO (exposure range) of the film.

Each one of the these three aspects affect how your final exposure looks like.

F Stop (size of the apeture opening) controls both the Depth of Field/Focus of your exposure as well the amount of light you're letting in. So the smaller your apeture (lower the F Stop) the deeper your field of focus is. But then you lose light and get dark under exposed images. The wider your apeture the "skinnier" your depth of field becomes, but you get more light, and can contribute to over exposing (blowing out) an exposure.

The shutter speed controls how long you expose the film to the light, which contributes greatly to capturing motion effectively. The slower your shutter speed, the more blurry something becomes when in motion. The higher your shutter speed, the clearer something appears when in motion. And of course the amount of time you expose the film also contributes to whether you get a dark under-exposed or bright over-overexposed image.

The ISO range of the film helps to compensate for these factors, especially light. The higher speed the film, the lower levels of light it requires. Thus ISO 100 is typically broad daylight speed film, and ISO 400 is typically used for indoor and lower lighting. The drawback with higher speed film is "noise" grain. The higher the speed for lower light, the grainier and grainier the image. Which is why a lot of photographers that specialize in night time photography use lower speed fillms like 200 to 400. They just mount the camera on a tripod, use a shutter release cable and pop the shutter open. They then leave the shutter open for like 4-8 hours and then click the shutter closes, and bingo amazing night image.

So let's say I want to take a wide landscape kind of photo of a beautiful little beach house on a cliff. As I look at this view with my eye I see this amazing image. The gorgeous sky with the sun that's setting, the amazing ocean, the house, the cool stuff on the lawn, and the gorgeous mailbox that's in the front yard way out in front of the house.

So I take the shot, and of course, it doesn't look like what I saw with my human eyeballs.

Each exposure has something lacking, the one with the low F STOP looks under exposed, The higher FSTOP gets the exposure right, but then the foreground and background are blurred.

So this is where HDR comes in.

There's an app called Photomatix that can take multiple exposures of the same image and then combine them into a single image. Then using the built in tone mapper, and down sampler, it creates an amazing 8bit/16bit image that couldn't be taken with a conventional camera.

So is there a way to do this with SD or HD video?

Like take three passes or shots of the same thing and then combine all of them into a single take?

So is there a way a ghetto/macguyver way to do HDR with an HVX and an editing suite like Edius or Vegas or FCP??
 
Well a lot would depend on what is moving in the shot. Let's say you wanted to take a 5 second shot like the wide landscape you mentioned with the house on the cliff etc.
If nothing in the frame was moving you could set the camera on a locked off tripod and use the remote. Record for 5 seconds, stop and change your settings, record for another 5 seconds, stop and change your settings, record another 5 seconds. Each time adjusting your settings for a different look. Now you could take all 3 clips into FCP and stack them, one on top of the other and play around with the opacity of each to get a blend of the 3 clips. You could also use masks with very soft edges. It may get you close to what you're looking for.
 
Hi, I just came across something about High Dynamic Range using PhotoShop,
and I'm sure another App like Corel's PhotoPaint,... etc .. can do the same thing
... BUT ... at this link though, it discusses ...
PhotoShop CS2 High Dynamic Range

I wonder if Vegas has something like this technique,... that would be nice,... ,

from what I've read, I think PhotoShop also handles Moving Video files too,
IF i'm not mistaken, so you may be able to put several video segments
through it to combine it for a stunning final video like the one you're interested in
... I'm not sure about PhotoShop and batch files ... to automate the process.

But it's a cheap start anyway ... :)

Ozduc makes a good point though, the scene couldn't move if you're using
ONE camera to take the several video shots you'd need with different
exposures.

so unless you've got several HVX's to take the same shot with the exposures
changed, you'd be limited to the shots you could make... I guess.
 
Last edited:
HDRI is used in video when you need to match move or do motion tracking. like when you need to shoot an empty street and place a 3D image such as a car or plane or whatever, this is where you need to get lighting right on the 3D CGI object to match the lighting surrounding it. So as your camera dollies the lighting on the object changes according to the shot.
 
Thanks for the replies guys.

Hey ozduc,
I've seen a very simple use of the layering using a mask in various films. It's usually an extreme close up on someone's face with like half their face in the shot on the right side of the frame, layered on top of someone standing far away. They usually cut the frame down the middle. The close up is usually done with a macro lens, and a long lens for the full shot. The thing is then you actually can make out the border where the two shots are mashed together. If they're good they'll blur the borders, but it's still obvious that it's a layer job. Now I don't have any experience on FCP but something tells me that the masking isn't going to be all that sophisticated. And what you're talking about would be more like a collage. The HDR images in the photo gallery blend the various exposures very close to seamlessly.

Hey Snapshot,
Yeah, I found out that CS2 does support HDR compositing through the bridge. But it doesn't have a built in tone mapper, so you have to mess with the s curve to get the image to look right when you down convert it to 8 bit or 16 bit. It does give you a little more tweaking room using the S-curve, but the Adobe Bridge interface doesn't support batches. Photomatix does batches, but it's intergrated tone mapper is a bit less tweakable. But I guess what you're talking about would be to export every frame of the footage, import it into Photoshop or Photomatix and then HDR them there, and then re-import them back into the editing app. I guess it could work, but it would be a very tedious proccess.

Hey Cache,
Yeah, my understanding is that HDR came from the CG industry. But it's being used now by conventional photographers to create uber exposures like the ones in the gallery I linked earlier. And that's exactly why I'm interested I guess. It creates a nearly CGI looking appearance to reality. I've been looking at more and more HDR photography and it gives a very surreal almost fantastical look.

Yeah, motion. A couple of my photography friends were saying the camera would have to be locked down and you couldn't move it at all.

So how about....

If it's just a wide establishing shot it would be okay.

I'd take my HVX w/M2 adapter with a wide angle SLR lens (with manual f ring), lock it down. Set my attached SLR lens to F2, shoot for 5 seconds. Stop down to F4 shoot for 5 seconds. Stop down to F8 shoot for 5 seconds. Reapeat up to F14.

Then I export all the frames from each take. Match up the frames from each take with their counterparts. Then use the HDR process in either Photomatix, or Photoshop CS2 with the bridge, using the exact same settings for all the frames. Then I'll have a single set of images, that I import back into the editing app (Vegas, FCP, Edius).

Now how to do it with movement??

Maybe create ghetto motion control?

Let's say I'm going to do a straight on horizontal pass from left to right. So basically the pass goes from west to east while the camera faces north if you will.

What if I take the tripod and lock it down onto a dolly with let's say a 20 yard track.

TAKE 1
I then measure off 5 yards from the left end of the track and use that as the starting point for the dolly's first take. I place a bright yellow pole in the scene to mark the left most area of the scene. Then I use a steady reliable motorized winch from the back of my truck to pull the dolly across the track at a constant speed. The SLR lens on my M2 adapter is set to F1.4

TAKE 2
I repeat but start from the left side 2.5 yards from the left end of the track and have the SLR lens set to F4

TAKE 3
Repeat but start from the left side 1.0 yards from the left end of the track and have the SLR lens set to F8

TAKE 4
Repeat once more from the beginning of the track and have the lens set to F14

Then export the frames from each take out as individual jpegs. Then trash all the frames that were taken before the camera passed the yellow pole say at around 5.5 yards from the left end of the track.

Photoshop/Photomatix the frames together into HDR, down sample, tone map, and re-import into editing app. (oh and it looks like Photomatix has a built in ghosting compensator to help remove ghosting from motion, but don't know how effective it is)

So what do you think? Sounds crazy doesn't it? Which is why I was asking if anyone knew of a better way? There has got to be isn't there? One that doesn't cost tens of thousands to rent motion control cameras?

Somebody help?

:)
 
Last edited:
Hi, kidgoldenarm,

I could have sworn PhotoShop supports importing the full motion video without having to import each separate frame ... hmmm ... I'll have to check that out again to make sure... as for the Tone Mapper and batch support ... have you looked at the latest version, or additions, updates? I don't have the program yet to know for sure, but I thought I read somewhere on the site (Adobe) that the latest version had those extra features now ... maybe I'm wrong, ... so many programs to DEMO, ... and so little time ... :D ... they're all starting to blend into one big DEMO in my mind now. :tongue:
 
Last edited:
The easiest way to do this would be using a 3CCD splitter, but instead of each color going to a separate CCD, all three colours go to each of the 3 CCDs, where 1 CCD is ND .0, second is ND .6 and third is ND .12. Then a little chip inside the camera splice these three identical videostreams to 1 and VOILA! 11 stops dynamic range. (this is basically the same stuff that HDR does to still images).

Of course manufacturers know this, but the resulting camera would be so good, no one would buy their more expensive cameras. So I guess we'll never see this solution.

I've already tried this, resulting in the clip here: www.neomedia.no/video/hdrvid.avi
 
Last edited:
where does one get a 3-CCD splitter? ... is it a dichroic prism used differently?

This sounds quite interesting


does the camera have to be opened? ... if so .. any mod pix's for camera's in general?
 
Last edited:
Check out debevec.org. This guy is a megabrain. I remember him talking about HDR almost 10 years ago.

A simpler solution than changing optical paths and splitters and such would simply be an oscillating electronic shutter. Have a camera that could refresh 72fps and expose the ccd for varying durations say 1/36th, 1/72, 1/288. You could acheive the same effect electronically. This would also eliminate any differences associated with the optical characteristics of lenses at different stops, specifically dof.
 
oh sorry... forgot to mention: I haven't modified my camera, it's just 3 different video streams with respectively ND.0, ND.6 and ND.12 merged into one. All blown-out whites and blacks are removed, so that only properly exposed video remains. The problem with this method is that it can only be used on still images, with the camera on a tripod (and without movement). So it's really not very useful for filmmaking.
 
You must log in or register to reply here.
Back
Top