(half baked knowledge incoming...)
S_Pfeiffer wrote:So I don't even think that things like color and chroma sampling like 4:2:2 and 4:4:4 even apply. I'm not sure about AE though.
If you debayer something from RAW to RGB (not YCrCb which I don't think any of the common programs is doing), then you have "equal" amounts of color information present in the channels (AFAIK 12 bits in the case of BMD RAW).
However, due to the design of the sensor, you usually have only HALF the resolution in R and B as opposed to G (which is one of the reasons that using a GREEN screen makes the most sense).
I'd say that keying off of a debayered RAW DNG frame sequence should generally be simpler and more accurate than off of a 4:2:2 ProRes stream.
BTW, I don't think stream vs. sequence has any bearing on 4:4:4 vs. 4:2:2 (see
http://en.wikipedia.org/wiki/Chroma_subsampling).
Side note: It would be interesting to know the actual pixel/photosite counts for each color of the BM(P)CC sensor.