Hi everyone,

I have been wondering lately if could be possible to utilize a technique from still digital photography to improve the dynamic range of an image beyond that of 35mm color film (not sure about B&W).

I have seen still photographers combine several images - one exposed for highlights, one for blacks, and one for midtones - into a perfectly exposed single image.

Granted, this can't work in video as it would in still photography as we can't easily take several different shots of the same thing and expect them to overlay perfectly.

(btw, it occured to me that you might be able to do this very simply with the color correction tools built into most NLE so this may be a completely stupid and roundabout way of accomplishing this).

Therefore, I was thinking that it would be easy enough to do this with one RAW image. If one exposed for the midtones (while maintaining highlight information) you could easily duplicate this footage several times and adjust each individual parameter. For instance you could bring up the midtones with a curves adjustment layer, select the highlights by inverting the image, applying a 1 pixel gaussian blur, and use this as a mask.

Am I just being stupid?

I had a similar idea but took it a different direction. Imagine a 3CCD camera where each CCD was instead recording the image in color but at different exposures. The camera would have a built in "zebra-like" feature in the sense that it could tell the parts of the image that where about to clip. Thoughs sections of the image would be omitted and replaced with the opposing video signal from the second CCD set at a lower exposure. Now apply the same concept to the the shadow side of the image and selectively replace the near blacks with the video from a third CCD thats set at a higher exposure. In theory you could increase your stop range by a amazing 3X! All the while no post processing would be needed and it could all be recorded to the same format as before because the resulting image going to tape would be NOT be the sum of three full signals but pieces from all three to adding up to only one signal.

Im taking orders starting 2006 ;-)

-Brett Erskine
www.CinematographerReels.com

That is an interesting idea Brett! Actually I think there is a technology used in some SLRs? that uses a similar idea:

Instead of three chips it contains a distribution of large and small image receptors with which it captures different areas of the image. You loose resolution (but then again the camera I know of that uses it is 6megapixels) but you gain dynamic range.

Well, there are logarithmic sensors available.
They work more or less like this:
every pixel has an added transistor that the higher the signal the pixel gets, the lower the gain applyed to that pixel.
The same applyes for shadows, the lower the signal the higher the gain...
Fillfactory's Lupa line work that way..

Some film scanners employ the multi exposure method when scanning print stocks.
Dalsa had the log exposure too, I remember.
Fuji is doing the different sized photo sites, right ?
-Les

See http://www.debevec.org for lots more on HDR imaging, including a free download of HDR Shop. Also do a search on these boards for HDR; I've posted about the topic several times in the past.

DEar Folks,
I use zone system to define tone when i work with DV, mostly with PANA DVX100 at cine-like, to myself i also stick to 5 stop range i have a conversion between the marker and zone
80% = VII
65% = VI
50% = V
35% = III
25% =II

so if i set up lighting i instentionally make a flat ratio to squezze all detail from hight light ( except hotspot) to shadow into 5 stop range, i have no problem to bring to any contrast or gamma at post.
just like the late Ansel Adams N+1 or N-1 with developer or paper.

the only problem is the orginial DV footage look very dark and dull, and mostly the client they worry a lot and even some directors insist it's too dark, but after all, a TV with higher contrast and brightest for client and a test for director will do.

I am not trained by TV/video, so i only try to make something out of DV or digital beta like film for a lower cost.
many fellow technician claim i mess up the system as they say "oh, it's too dark, too low contrast."
but i am shooting a film with DV, i also don;t do white balance i just set like 5600K or 3200K. this wil give u a unique colour to deal with at colour correction, one set will fit most of the scene.

i recently read and article that even 2 stop under on DVX100, the gamma curve can be bring back to normal, so this is a way to preserve the detail on highlight.

Think FILM, not video.
good luck
yamamoto
DP

well, Yamamoto, I'm a video technician (for video to film works)and I work exactly the way you describe, and I suffer the same things as you do...and I'm in Argentina!!!

So it looks all around the world cine people think the same way :)

The problem with under exposing video is the noticeable increase in noise. when the scenes are lightened to proper levels or tonality. Under exposing also reduces the amount of detail your image will have and can cause banding.

Ansel Adams created his now well know Zone System in the 1930's. He never intended for it to be used with video. Video chips have a different response curve than film and Adam's methods really aren't applicable.

well, it works perfect for me anyway :)

The zone system is a system dealing with Black & white film and developer and paper of it's time, but once you understand it, it's a scale like music, so a simple test or search of the curve will give you the idea easily to work with.

I actually did a feature film with panasonic tiny DV camera, dx-100
i choose to shoot almost everything just at sunrise or about sunset and maniputaed at post to deal with contrast and garde the colour i want. almost no one believe it's DV. they all ask is it film.

noise is the issue that's why i go flat to avoid too dark and bring up from grave.
Tiffen has a filter to soften the contrast ( ultra contrast)

also if ur camera allows to use some lens from earlier which is single coated, e.g i use old nikon lens on mini35 it's work ok.

The cine-like gamma from DVX100 is a trend to camera manufaturer to look at the more dynamic range to response to the film makers who want to use dv as an effective medium to shoot. it helps me to preserve the detail. just like negtive does.

however my friend who used to runs a cable Tv hates film gamma and always prefer 4:3 sharp edge enhance and high contrast.
so it's a matter of taste too. he thinks this is a mistake to make DARK and DULL video.

YAMAOTO

Indeed like Juan writes, Fillfactory (Belgian spin-ff from IMEC) has an active pixel CMOS imager technology which in the future will make it possible for consumer electronic imagers to get dynamic ranges far beyond (100db and more) the film lattide possibilities. Each individual pixel holds an auto gain reduction feature resulting in a real DR compression of scene DR. On the other hand pro cams because of the better (larger...) CCD chips, the more precise initial AD conversion (13bits and more)and signal processing (knee processing..) can go up to 80db DR without excessive noise or banding. Consumer cams seldon go beyond 50 db for an acceptable picture

Back to Brett's idea - why not use a three chip block but have an ND filter pattern on each chip? One clear, then ND3 ND6 ND9.
You'd have to record at a higher data rate to absorb the HDR image, and you'd be interpolating down to a quarter of your actual resolution per colour. But assuming the image wasn't insanely contrasty it should look pretty good.

... the cheapest and most effective way is to use a rotateable gradient ND filter.

I use a ND gradient (-2 stop to clear) which i always carry with me

I also have a gradient mask for it - easily made by shooting a white card with the filter on.

the gradient mask combined with a garbage matte is used in post to pretty much sort out any extreme contrasty conditions

Canon already has a patent on a CCD/CMOS variable DR imaging chip from the early '90's. If I remember correctly, it uses variable CMY dichroic filters to adjust the exposure.

Jeff, was this an intra scene variable DR system?

I haven't forgotten your post Andre, I've been looking for the paper I read on the patent. But as I recall it was for intra scene improvement of DR. I'm still looking though.

I was using recently on a shoot a technique with employed two exposures with the camera locked off. The second exposure was for highlights only with the in-camera ND filter. The images were blended later in AfterEfects and looked stunning.... Would have been nice to have a motion control rig with this as well...
Following this why not having a second chip installed on a prism with only an ND filter attached? I am not a video technician but just thinking out loud: the signal from the second "ND-ed" chip could be mixed (overlaid) with the main one after some filtering is applied to it, similar to audio LowPass or HighPass filters?

What a great idea! ;-)

I was using recently on a shoot a technique with employed two exposures with the camera locked off. The second exposure was for highlights only with the in-camera ND filter. The images were blended later in AfterEfects and looked stunning.... Would have been nice to have a motion control rig with this as well...

This is an interesting idea -- someone showed me HDR stills recently, compiled from a blend of High, Low and Balanced exposures -- each shot he took was comprised of those three settings, then the software allowed him to use the interpolations to do an array of things like dodge/burn accurately...

For video, this would take three separate CCD and lense systems, recording streams simultaneously :/ Or...perhaps a 90fps recorder that would arrange the high/mid/low exposures across every third frame. Either would take a bit of work...

- jim

Am I missing something obvious in the idea of a bayer style ND filter pattern on all 3 chips of a 3 chip system?
Of course the beamsplitter-lens arrangement gets more complicated but the system exists in every decent camcorder.

The grid would look like:

----------
0 | +1
----------
+2 | -1
----------

Surely a decent HDRI could be interpolated? There would be less data lost than in a bayer pattern because of the exposure overlaps.

In theory this could be done in conjunction with a bayer filter on a single chip system, as long as you had the spare resolution, like a 2k chip for SD output. Or am I missing something?

I was thinking at only two chips with a three stops ND filter. This could push the dynamic range to ~11-12 stops. But Dan's idea is interesting too. Why not use a single Bayer chip config, similar to the one in Fuji digital cameras (a honeycomb config with chips of different sizes - the smaller ones to be activated only on higher levels of light...) Don't have a URL handy but it's worth a look

The honeycomb structure is really rather interesting. Quite ingenious actually! I wonder why we haven't seen this sort of thing introduced for video yet though... even high end cinema cameras don't use this (As far as I know). Perhaps it has something to do with the patent?

The images were blended later in AfterEfects and looked stunning.... Would have been nice to have a motion control rig with this as well...

How did this work exactly? Was it simply a still shot of, say, nature or were there actors on screen? I imagine it would be incredibly difficult to match two sets of footage in post if actors were moving about. Even with the camera locked off. Hence the reason it would be so nice to have RAW video. You could pull the necessary highlight clip and shadow clip from a single shot/take.

<<<-- Originally posted by Aaron Shaw :
How did this work exactly? Was it simply a still shot of, say, nature or were there actors on screen? I imagine it would be incredibly difficult to match two sets of footage in post if actors were moving about. Even with the camera locked off. Hence the reason it would be so nice to have RAW video. You could pull the necessary highlight clip and shadow clip from a single shot/take. -->>>

Aaron,
There were locked off shots. The foreground action was recorded with the proper exposure (~1/3 - 1/2 stop under). The second exposure employed one of the ND filters in the camera. We did not need to repeat the foreground action because that area will fall into the mask space.
The shots were layered in AE masked and blended. The advantage is that the mask can be "painted" only in the area you wish to.
I don't think that a RAW file will support such a dynamic range. With the method described above, you could claim almost 15 stops.
Just try to shoot RAW with a digital camera and experiment. The main advantage would be for the colour correction where yo can choose in post the colour space and minor luminance variations.

Here is a link for more info about the Fuji sensors http://www.dpreview.com/news/0301/03012202fujisuperccdsr.asp and their photosites structue.

Ah yes of course! How simple! Thanks for the clarification. I never would have thought about doing it that way. Interesting... I'll have to try that sometime. Masking can be a real pain I assume though.

15 stops being the dynamic range of the combined footage not RAW file I presume? I know that some still photographers shoot one raw file when doing this sort of thing and others shoot two consecutive shots. The two separate shots looks somewhat better but the RAW file isn't half bad either, though significantly less useful.

Thanks for the resonse!

I shoot almost all the time RAW stills using two exposures and layer them in Photoshop. I will post a link to see some samples.
It is very easy. The mask can be painted with broad soft brushes or any size and shape you want. You can also make an "action" and automate the process.
I will post a link to see some samples.

Great! I'd love to see some examples. Do you have some video clips from the shoot you could post? I can host files if you would like.

I have rushes/dailies on DVD. I'll try to convert them to a QT file. I will post some examples with Photoshop files as well.

<<<-- Originally posted by Jeff Donald : Canon already has a patent on a CCD/CMOS variable DR imaging chip from the early '90's. If I remember correctly, it uses variable CMY dichroic filters to adjust the exposure. -->>>

Interesting, thanks.

Jeff, I have also been looking at doing a range compression trick commercialy. This Canon technique, can you tell me more about it, just in case it is the same?