Both C300 and Scarlet have CMOS Bayer Pattern Colour Filter Arrays. The Scarlet's is 5120x2700, albeit with a cropped region of 4096x2160 typically used for movie capture, all photosites being read individually. The C300 is 3840x2160. It's not sure at this stage if all the photosites are read individually on the C300 or not - it could be that the pair of greens are read individually and then combined, or they could be combined as they are read out.
As the C300 produces a 1920x1080 direct image from the Bayer pattern, we must compare this to how the conventional method would be, which is to produce a full demosaic of the image, then downsample to 1920x1080 with proper filtering. I've attached an image comparing these two alternative methods.
The "quick extraction" method as we think is used by the C300 shows chroma moire because no attempt is made to interpolate to align the red and blue photosites with the greens. It also shows some aliasing as you've produced a smaller image without using a downsampling filter.
The full extraction method as you'd use with the REDs shows a smoother image, less aliasing, and no chroma moire. It's also a tad sharper.
I produced the example demonstration quickly, but when I get the chance I'm going to code up a fuller demonstration, and also one that will feed in more high frequency detail to show what happens as both systems are fed image data that would produce more severe aliasing.
Of course, what we need to do is get the C300 on a test chart and see how it performs. It could be that they use a much stronger OLPF (with the increased loss of resolution), or it could be that they use their typical DSLR OLPF and we'll get a similar result as to what we get with sRAW. I've attached an example from sRAW that I did with the 1DmkIII, showing that although chroma moire is reduced, luma aliasing is vastly increased, giving additional artifacts compared to a full demosaic.
In practise, the RED's perform exceptionally well with regards to aliasing and moire. Not perfect as that'd be nigh on impossible to achieve and still have image sharpness, but certainly the best of any movie camera I've measured and I've had to measure quite a few.
But what are the benefits of a quick extraction approach? Simple - it needs next to no hardware to achieve, and that makes it both quick and cheap to implement. And the drawbacks are that it impedes image quality.
Coming from a stills background, aliasing is not the issue it is with movie capture. On a still, if you have an issue you can perhaps go and paint it out by hand, but you can't do that on motion easily. Aliases also move in the opposite direction to that of the motion in the scene, causing motion adaptive codecs like MPEG2 extra work to compress the image. And motion makes aliasing more easily visible than in a still.
Graeme
