Hdr

[Donald Harvey]

So I love photography as much as I do filming, and I came over and interesting idea. I bet it has already been done, or done in some way but it would be something nice to know about.

In photography you can do an effect called HDR, which most of you probably know about. But what you do is take a picture with a low exposure, mid exposure, and high exposure. Melt them all together using a HDR program. The results are crazy, you an see some beautiful things that make a picture much better.

When you are taking these three exposed shots you need a very steady camera, so using a tripod is necessary.

Are we able to get this sort of effect for video...I know that you cannot have the actor do exactly the same thing for three different shots, and if you could it would be very very time consuming as well as money consuming. But another way you can approach this in photography is by taking a picture and lowering the exposure and highering the exposure in post. You get the same effect but not as true colors. But it still looks amazing. Are we able to do this with raw video?

Let me know what you think and if it is possible, obviosuly it would take a lot of rendering but the product would be very good looking.

 

[Ryan Patrick O'Hara]

'Bracketing' is an advantage still photographers have over cinematographers. You are basically describing a way to get very high dynamic range by mixing bracketed exposures. Unfortunately when the pictures are 'moving' these techniques of blending multiple exposures becomes very challenging, if not impossible.

Although I see what you are saying and this technique could have its place and time.... even despite how hard this would be to do, there is an even larger reason why this isn't very possible. When discussing 'normal' film making circumstances, it's just not feasible. Film stock is expensive. Crew wages are expensive. Time is money. Shooting scenes three times with different exposures would simply elongate the schedule and cost huge amounts of money. Plus, not to mention all the hours (and extra money lost) in post production dedicated to mixing these images together.

In the end, this is one of the reasons a Director of Photography is hired. He oversees the lighting, and will be able to know if something does not work, and then make it work. By using light, filters, and lighting control, the DP can make the image look good in one exposure.

 

[Thebes]

I think its important to differentiate tone mapping and HDR. They go hand in hand, but they are two different things and most people act as if they are one in the same.

Out of curiosity I tonemapped some video from my HV20... I do realize that its not a HDR image. I exported frames out and ran them through a batch job in photoshop- open file, convert file to 32 bits, convert file to 8 bits with local contrast adjustments, save file. I took one image from the frames which I thought was representative of the scene and used my settings for it with the batch. Pretty simple, IIRC it took my Core2Duo 2.6ghz 2gb XP Pro a bit less than a minute per second of 24p video to process the batch. Filesize was huge, of course, with tiffs. It worked pretty well for mess around stuff and gave the look one would expect with single jpg "hdr" photos processed in photoshop.

 

[USLatin]

I guess actual HDR may happen the day someone figures out a way to get the 180 degree look with ultra fast exposures without the need for ultra-slo motion lighting... meaning a sensor that were so sensitive and had fast enough refresh reset times to shoot 48fps with each two frames exposed at supper slow motion rates super close together... the problem is that you would get the small shutter angle look... though I guess that may be just fine and dandy for some.

Or, even more interestinly, you could split the light into two sensors within one body... but that would mean you'd be using half the light for each sensor and you would need more light... unless there were a way to get more light in though the taking lens... Ryan, is this what evoked the "IF not impossible"?

I shouldn't be getting myself started on this...

 

[Ryan Patrick O'Hara]

The 'impossible' comment is targeted at how incredibly difficult it would be to mix bracketed shots together in post. The 'impossible' part would be when the camera moves or actors are in the frame performing. It is possible, technically, but since it will never be the same way twice, you would need someone very talented to roto the entire scene, frame by frame.

In the end, shooting triple the amount of shots on your already established coverage plus the post work would make this 'impossible' from any producers lips.

Although I still can't think of a special circumstance where this would be necessary, I do however enjoy USLatin's train of thought about capturing two exposures SIMULTANEOUSLY.

If you know how a reflex film camera works, the thought of using a mirror shutter to direct half of the light into another capture medium (instead of the eye piece) is very simple... in theory.

This way you can get either the high or low end of the specturm. For instance expose normally, and have the mirror shutter throw the light through a filter causing the other exposure to be darker. Or you could overexpose the original exposure and run the footage through a filter causing the second exposure to be the correct amount of light.

Now, I'm not a video guy, so I'm not sure how they (ccd's) exactly work, but I figure that they would not loose a stop of sensitivity because the light is being directed to another capturing chip. The camera would simply need to ditch the electronic shutter and have a manual shutter.

I'm sure someone could develop a chip that could alternate exposures constantly on a single ccd... but I doubt that technology is here. My example is with a physical shutter because it's easier to picture in the head and more practical at this time.

Very cool to think about theoretically, but I still don't see a market for it.

All and all, I think most of my negatives about this process were eliminated by the simultaneous idea Latin brought up.

:beer::dankk2:

 

[USLatin]

The digital world works the same way as film does. The "only" difference is that you have a sensor that stays in place and instead of traveling film you have refresh reset times which are getting WAY faster. Red One for example. Scarlet is poised to shoot 120fps continuously. But I would suppose that you would need even faster fps capabilities than that to get two exposures close enough so there is practically no discrepancies that would cause blur when merged.

What I am thinking would be best is to use film methods and do it on the front end with some fancy optics. One lens takes in the light, a mirror reflects (somehow!?!) half the light down a different tube... is that possible? I imagine it would be much like the optical viewfinder developed for Red. I can't remember the company but they make a front face plate for Red One that does just that I suppose. You might know better having worked with film cameras whether this is possible. But then the chips/gates inside are permanently offset by means of a very nice ND on one only one them so that there isn't other differences in the image from different ISO/gain/stock. Or you could allow for an opening where you could drop in a series of different NDs... or easier yet have a variety of NDs in an internal wheel so that all the different NDs end up in the same placement.

I guess it could also be a mount where you could get two cameras of your choosing...

Then you light for the wider range. Though that would somewhat defeat the purpose cause if you are lighting and you are not outputting to something with such wide contrast capabilities such as a DVD. Then you might as well save yourself the hassle. Unless you are using available light... in which case it would be very handy for something like mid-day sun to get the highlights separately and the shadows on the other... we could make video beat film in DNR today! :beer:

You get a ton more info and you squeeze it and twist it in post to your liking. Post would be a little trickier, much like RAW with an added step and double the files (double the fun).

The only problem would be that your shadows would be from a camera that is only getting half the light it would by itself... and I guess that might be why it hasn't been done... you sorta shoot yourself in a foot half way cause you get more DNR but you end up lowering your low light sensitivity.

 

[Ryan Patrick O'Hara]

The only problem would be that your shadows would be from a camera that is only getting half the light it would by itself... and I guess that might be why it hasn't been done... you sorta shoot yourself in a foot half way cause you get more DNR but you end up lowering your low light sensitivity.

I don't see how you are loosing half the light. You're actually just putting what would have been lost (from the mirror shutter to the eye piece) to work. (speaking in film terms. although I do believe many high end digital cinema cameras have mechanical shutters)

This is all theoretical. Just to get the image as sharp as the first exposure by relaying with a mirror shutter or etc, would prove a small challenge. There is not a way, I know of, to split an image in the lens. A mirror shutter or prism, would be the best bet. I like the mirror shutter option because it relays the light that otherwise goes to waste NOT hitting the original exposure. If you were to shoot with a mechanical shutter, half the light never reaches the original exposure anyhow, only now instead of being sent up the pipeline to the eye piece, it can go to another, synced film plane. In theory quite possible.

 

[USLatin]

OH!!! So you have something like a spinning mirror of sorts or somehow make the other half of the 180 degree shutter not get blocked by the rolling shutter but somehow pipe that light into the other camera... I think I see that's what you are saying... but then the exposures would happen at slightly different times creating a little blur in the merged frames.

I was thinking of a mirror that bounced half the light and let the other half right though it. I thought that's how the EVF for Red One worked.

 

[Ryan Patrick O'Hara]

OH!!! So you have something like a spinning mirror of sorts or somehow make the other half of the 180 degree shutter not get blocked by the rolling shutter but somehow pipe that light into the other camera... I think I see that's what you are saying... but then the exposures would happen at slightly different times creating a little blur in the merged frames.

I was thinking of a mirror that bounced half the light and let the other half right though it. I thought that's how the EVF for Red One worked.

I'm not sure how the Electrical View Finder on a RED works. I'm guessing it's just a preview "video tap" off the same camera chip. I doubt it has to pass through a mirror, because that would probably loose a good deal of light.

Also you just went to rolling shutter which is different territory altogether. I don't know how this would work with rolling shutter as the frame does not expose in one even field. I'm going talk about film exposure, global shutters on CCD's and future CMOS sensors. Rolling shutter is just a pain in the ass.

However a digital global shutter synced with a rotating mechanical mirror shutter (when digital is taking a frame, mirror is out of the way, vice versa) could in fact redirect the unused light (frames between frames) to another medium. This would have very minimal differences between frame 1 and 1a, but fast motion, etc would have differences that would have to be sorted out in post somehow. The difference mind you is equivalent to a difference in the frame of a camera capturing 48fps. (In all seriousness, you are capturing 48 distinct frames per second by an off synced mirror relay to two capture mediums)

This means the two recording ccd's or film planes would need to be off synced.

A prism is the best bet for actual simultaneous frame capturing. This would loose a stop of light by directing the light at two synced chips or film planes at the exact same time.

 

[Ryan Patrick O'Hara]

After all this talk, if they (Panasonic) made two HVX CCD chips record off of an image split by prism, and had the 60fps chips off synced to record frames between the exposures of the other, could we see in the 'dead' shutter zones and get more fluent slow motion (120 unique frames per second theoretical)

I'm sure a tech guy who knows how the chips really work could shoot this down. This is why I'm a film guy. Mechanics are complex but easier to understand... no jibberish.

 

[USLatin]

I'm not sure how the Electrical View Finder on a RED works. I'm guessing it's just a preview "video tap" off the same camera chip. I doubt it has to pass through a mirror, because that would probably loose a good deal of light.

Also you just went to rolling shutter which is different territory altogether. I don't know how this would work with rolling shutter as the frame does not expose in one even field. I'm going talk about film exposure, global shutters on CCD's and future CMOS sensors. Rolling shutter is just a pain in the ass.

I think it is an optical mirror but I really don't know about that kind of stuff...

I meant to say spinning shutter... as in mechanical film shutter

 

[USLatin]

I finished reading your posts and you are dead on! 3CCD prisms! Only you use 6 CCDs and three prisms. That's it. Or one prism for two CMOS sensors (and make those Red sensors!).

Duh!

I wonder why this isn't done? Yes, it would be half the light hitting the sensors possibly less... but I bet we are getting to the point where sensors will be sensitive enough real soon based on the pace that Red is setting.

It would be SWEET to see a double-magazine camera with a prism! YEA BABY!

 

[filmjeff]

Or just do this...

http://grail.cs.washington.edu/projects/videoenhancement/videoEnhancement.htm

Well, whenever that technology is available.

 

[Drew Ott]

I think if video cameras shot in full raw like some still cameras do, HDR-like images could be obtainable. I've been doing single-imagine HDR by saving 3 copies of the same image at different exposures and treating them like bracketed shots. Works great.

It seems like it could be done somewhat easily in video if the codec wasn't so compressed.

 

[USLatin]

I think if video cameras shot in full raw like some still cameras do, HDR-like images could be obtainable. I've been doing single-imagine HDR by saving 3 copies of the same image at different exposures and treating them like bracketed shots. Works great.

It seems like it could be done somewhat easily in video if the codec wasn't so compressed.

That sounds like Red One to me.

 

[USLatin]

Or just do this...

http://grail.cs.washington.edu/projects/videoenhancement/videoEnhancement.htm

Well, whenever that technology is available.

That's some real crazy stuff... I wonder how much power/time it takes and how it looks at full resolution... how hard it is to get the pictures to work, and how much time and effort you need to put into taking the stills... mainly I wonder if it is any good at full resolution.

Keep in mind that to achieve HDR with this method you would have to cover everything in HDR meaning taking different exposures of everything and making the HDRs before doing this process...

Finally there is no way this would work in an automated way with people... and it might not be automated at all...

then from a production stand-point you would have to shoot all kinds of pictures for every take which would tie up the set, and and the actors making everything take twice as long which would, by itself, close 99.99999999% of every producer's office doors in your face.



But the biggest problem would be doing it with people... or anything that moves... the lips make it completely impossible...


yea... there is no way in hell.



But that's some neat stuff for sure :thumbup: