Traditional Film and HDRI

[rendermaniac]

Taking HDRI sequences with a film camera is perfectly possible, but there are a lot of extra steps to getting yuor final image. You need to develop it, scan it and stabilise the images before you can even think about combining them. Registration can be a real pain (although a one point Shake track tends to be OK).

With a digital camera you can use the SDK to take all the photos for you and combine them in minutes - especially with the Canon range of digital cameras - see Sean O'Malley's for some excellent, free software for doing this http://www2.cs.uh.edu/~somalley/hdri_images.html.

Plus they will definitely register properly and the images often contain exposure information.

Film does (currently) have a wider native dynamic range than digital so theoretically you need less shots (as long as your scanner can cover that range), but digital is far more convenient.

Simon

[lehthanis]

You need to develop it, scan it and stabilise the images before you can even think about combining them. Registration can be a real pain (although a one point Shake track tends to be OK).
Thanx!

I don't think I get what ya mean when you refer to registration and shake track? Sorry...I'm new to creating HDRI and I'm not a photographer, my wife is but hobby only.

[Andrew W]

I don't think I get what ya mean when you refer to registration and shake track? Sorry...I'm new to creating HDRI and I'm not a photographer, my wife is but hobby only.
Registration is the act of getting 2 images to line up on top of one another. Digital cameras do this automagically because their capture array stays put and provided you don't move the camera between exposures they'll line up when you import them into your computer.

Film gets advanced through the camera, it warps and stretches slightly during processing, then you have to scan your negs (or prints) so you have to make sure that they line up afterwards as any of the above process can intoduce discrepancies between exposures. This is why you have to register your images. Some compositing packages (i.e. Apple's Shake) have stablising algorithms that will do that, as rendermaniac mentioned, but if you don't have access to such software you can do the same thing in Photoshop by putting each image on its own layer, making it semi transparent and moving 'em arround till they all line up. Then make each layer fully opaque again and export each layer as a seperate image. Et voila, you're ready to put in HDRSHop or equivalent.

Best,

Andrew

[lehthanis]

Ahhh ok...simple concepts, complex terms :D just kidding. Thanx. I gotcha. I'm trying to convince my wife to get me a nice Nikon D70, but thats not as easy a battle as it should be. That would solve my problem. till then its film and scanners for me.

[tweeeker]

In addition - I'd recommend just going for it and not worrying too much about the details to begin with. Don't forget that to approximate diffuse reflections your gonna have to blur your hdr image by a large amount (usually 30% - 50% of the overall width). That allows for quite a lot of error. If you also want to use your images for high quality specular reflections then thats a different matter.

Good luck

T

[rendermaniac]

Unfortunately you cannot forget registration completely - even it is going to be blurred to hell and back. The dodgy red dots in my first attempts at HDRI a long time ago (done on film) at http://www.rendermania.com/HDRI/ are mainly because of registration issues. If HDRshop (or hdrgen at the time I think) does not have the same information at each pixel then it can confuse it a lot.

I would recommend trying some indoor scenes rather than having moving clouds and trees like I did!

Not to put you off though - the wasted film is a learning experience ;)

Simon

[lehthanis]

Excellent advice, thank you...

Anyone ever experimented with black and white HDRI? I ask because my wife currently has B&W film loaded, and I'm thinkign of firing off a few sequences...

then figuring out how to render in true black & white...shoudl be a simple matter right?

[rendermaniac]

Black and white should work - the combination process only needs to know the exposure difference rather than colour information. This will get you back the brightness of the light from the original scene. You may even be able to use this with a low dynamic range map to boost the colours to more realistic levels. (You won't deal with any colour shift that your camera introduces - although I doubt this is a big problem).

If you want to be a smartass you could shoot your range of exposures 3 times with red, green and blue colour filters and try and get the colour information back. Of course you'll have 3 times as many images to process!

Simon

[squidfrog]

Hey Simon--small world. :p

Incidentally, regarding B&W HDRIs, I've done a couple, but only in digital. I've also received examples from others who've produced B&W HDRIs and no problems were reported. In fact, in some ways they seem to contain more spatial detail than color images (in addition to the additional "spectral detail" you get from the HDRI process itself), but this may just be the channel-blending effect. (Similar to the pan-sharpening done for satellite data.)

Regarding registration: Simon's absolutely correct in stressing the importance of this. From what I read (somewhere), the "cathedral" image that's almost ubiquitous in HDRI research was joined from a set of film images, all registered together, and subsequently cropped from the originals due to the discrepencies in camera position between shots. Registration between different-exposure shots sounds tricky, though, especially at the low and high exposure ends. (And I'm not entirely sure if any non-academic software exists for image registration like this to begin with!) So any kind of camera interface that lets the exposure be set and the shutter to be released without touching the camera is definitely recommended--registration won't be a worry at all then. Just motion of the subject. ;)

Sean O'

[rendermaniac]

Hi Sean

Guess someones been checking their server logs ;) I guess that makes sense for black and white if you convert to black and white and then combine them. I suppose if you had a CCD that just measured luminosity then it could be much higher resolution as you'd have 3x (or whatever Bayer patterns work out to) more information. It would be interesting t osee if a Foveon CCD was any easier to use.

I don't know if black and white film would be better than colour film. I guess the three layers of film might focus at different heights which a scanner might blur.

Simon

[CGTalk Moderation]

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.