Hi all - wasn't quite sure what category to put this in but thought I'd pose it as a general question - since you all are certainly the braintrust behind the inner workings of Canons.
I'm working on a stop motion animation project being shot with a 5D, which I realize has limited chdk support but I think you all might have some ideas. The problem I'm running into is rearing it's ugly head when shooting greenscreen shots (you know, think "news weatherman" for those non-visual effects savvy people). Although the green channel of the RAW images seems to be fairly well put together, the Red and Blue exhibit weird artifacts that look like they are being subsampled and possibly sharpened along the edges. This makes the typical processes we use to key the greenscreen VERY difficult and messy.
I remember reading somewhere that the sensor favors the green channel, which makes perfect sense since our eyes are most sensitive to green. But in doing so it, you know, seems to sample red and blue at almost 1/2 as much data. it looks similar to what DV video is doing to optimize the bandwidth. (what is that, 4:1:1? whatever the compression is)
so here's what I'm wondering - since this is a stop motion project and is a very controlled environment, bracketing 3 shots in a row is certainly possible, and almost easy! we certainly do it for HDR every once in a while if the shot dictates a need for it.
I'm just wondering if it's conceivable and within the limits of hardware + firmware to capture all 3 channels using the better "green" channel of the sensor. So Red and Blue would be captured with as much resolution and clarity. And then combine the channels later. so think of it as - I know it's more complicated than this, but if the sensor captures light as RGB, could it grab RGB, GRB and RBG in 3 shots and then we just slot the green channels into the correct channels as a post process?
i mean, never did I think I'd ask if I could capture a red channel using a green channel, but you guys see where I'm going here and the reasons for it.
actually, because of the way they filter light to the different sensors, that's probably all hard coded and can't be changed around. but I wonder if you could shift which channel is being subsampled in the firmware. so instead of red and blue being downsampled, I wonder if you could bracket the favoring of R, then G, then B?
whaddya think? possible?