HDR Photography

[Michael Lindsay]

Hi David

Why would the photosites intended for the highlights have to be the same size as the standard ones. I would have imagined (excuse me if my imagination is none too technical) it would make more sense for them to be much smaller. One so they collect less light. Two because 'I' would be happy for them to have less resolution than those designed for the main part of the image. I suspect, like speakers, the crossover might be the tricky part and would be best placed high.

I'd be super delighted with 3-4 more stops of less accurate picture that kicked in where the digital sensor hard clips..

My experience of film suggests that its high highlights are compressed and noisy. . Most easily seen if, after exposing, you have to use them lover down.

If one was shooting in a studio with complete control they ( the highlight photosites) could be ignored and only the main photosites could make up the picture...

wht do you think..?

regards

Michael

PS Maybe if the main photosites where rounder (I believe there might be other advantages) they could place some little highlight diamond photsites in the gaps capturing the highlights. This might help stop a reduction in resolution.

 

[Michael Lindsay]

As for red: A 1600th of a second read out on the tail of a exposure (normally 48th or so) would give a, only slightly lagging, 4-5 stop highlight detail fill in that may be acceptable for some photography...

If you did one before the exposure you could do some pixel vector tracking so to place the detail where it is suposed to be (with temporaly challenging material)..

I assumed they were up to something like this when they they announced better DR at 25fps or bellow on the new board...


Michael

 

[Yannick Hagman]

Thanks

 

[Tom Lowe]

David's ideas are interesting, but I don't think the major hurdle is capturing two or three exposures of the same frame.

The problem comes when you, as purefilm put it, "Open Photoshop - open your 3 files - file - automate - Merge to HDR."

The automated merging of the image can be inconsistent from frame to frame, so the resulting video looks chaotic and flickery. The more movement in the frame, the more awful the HDR merge will look, from what I have seen.

 

[Noah Kadner]

I wonder if it would actually even be neccesary to create double the photosites. What if you just created some sort of special sensor that measured the overall level of light and analyzed it. You could then create a luminosity channel that would go in addition to the RGB. Then that is used in post to put back in additional latitude to what was actually photographed. Again we're talking big math but let's assume that at some point in the future, resolution will hit a zenith and we'll start looking to things like improved dynamic range as new camera desires.

Noah

 

[David Mullen ASC]

My experience of film suggests that its high highlights are compressed and noisy. .

That's usually an electronic artifact on some telecines due to trying to push more light or boost the signal through the darkest parts of the negative (the highlights.) Some telecines don't have this problem.

There is no particular reason in color negative for the brightest areas to have more grain than the darker areas. Noise is an electronic quality. And certainly the greater amount of detail in the brightest areas before they roll-off into pure white is more natural-looking than the more abrupt clip point with digital photography.

 

[Bruce Allen]

NoahK, one question in relation to your picture example - why do two RAW exposures? Why not just output the whole thing as a 16-bit TIFF (making sure to preserve the whole captured range) and then just do everything via power windows in the post package of your choice? PS your avatar is cute.

Why would the photosites intended for the highlights have to be the same size as the standard ones. I would have imagined (excuse me if my imagination is none too technical) it would make more sense for them to be much smaller. One so they collect less light.

You are aware that the Fuji still cams use exactly this system, right?

This (and many permutations involving different color filter array alternatives to bayer sensors, etc) is in a few other discussions:
http://www.reduser.net/forum/showthread.php?t=5196
http://www.reduser.net/forum/showthread.php?t=4982&page=2
http://www.reduser.net/forum/showthread.php?t=3126
http://www.reduser.net/forum/showthread.php?t=2853

RE: getting HDR out of a single Raw file, the question is "how high is High" in High Dynamic Range? Truly high dynamic range (eg enough info to do convincing HDR lighting in a 3D package) requires multiple exposures currently and / or a bit of hand tweaking.

As a post guy I am skeptical about the idea of shooting 48fps with an alternating ND wheel and using optical flow to make it all line up. Sometimes it just won't work. At the very least, you'll want a system to do that optical flow lining-up in real time in the monitor order to preview whether your footage will work.

When we get to > 200fps then maybe we'll really be talking. Of course if there were no penalty for doing sensor readouts, then we could drift off into an esoteric realm where we do 5000 low-quality sensor readouts a second and then combine them (giving us HDR and high speed at the same time)... but there are current practical limitations to this method obviously.

Finally, you can apply local contrast enhancement / Photomatix-style looks to anything, even low dynamic range images. Sometimes it looks great, sometimes not. Again, power windows are an example of local contrast enhancement done by hand.

Bruce Allen
www.boacinema.com

 

[Noah Kadner]

NoahK, one question in relation to your picture example - why do two RAW exposures? Why not just output the whole thing as a 16-bit TIFF (making sure to preserve the whole captured range) and then just do everything via power windows in the post package of your choice? PS your avatar is cute.

I would have to ask Graeme but I imagine the dynamic range you could get from dialing exposure way up and then down in Red Alert vs the full latitude in a TIFF file are not identical. But then again maybe they are. And thanks, I was very lucky with my good avatar looks.

Noah

 

[Michael Lindsay]

That's usually an electronic artifact on some telecines due to trying to push more light or boost the signal through the darkest parts of the negative (the highlights.) Some telecines don't have this problem.

My experience is limited as I only ever get to see my film after some electronic process. I think I've been too slow in my career to ever traditionally time film and see it in print...

What Telecine machine do you find have this problem?

Cintel machine seem to.. And Spirit either 2k scan's or traditional TK seem to as well... This is why I assume its the film..

Maybe I might look at some stills film on a negscanner and see what I find...