rick.lang wrote:The most important advantage shooting 2K for HD deliverables may be the image stabilization it affords to remove those little flaws and help smooth camera movement. May mean no loss of sharpness if your micro jitters are no more than a few percent of your frame.
Sent from my iPhone using Tapatalk
Sort of. Images that have transformations applied (i.e) stabilization, reframing, tracking, etc. apply a type of image filtering math to reconstruct the new image. Some of those soften the image, or some of those apply sharpening, but either way if you take an image that's 2048x1152 and resize it to 1080p, you get an image that is objectively worse than a 100% 1080p crop. Even doing things like stabilization, the difference in quality is minimal. In order to get the same quality as a 1080p source, you need to have enough oversampling to negate the quality loss, which makes a 1.5x oversample at 2880x1620 the perfect sweet spot of data rate, cpu consumption, and quality. I don't see any reason why the camera can't handle this in short bursts.
In fusion there is a filter type called "nearest neighbor." I don't think there is an equivalent in Nuke that I have seen and it doesn't average the pixels together so you get a very sharp image, but it's unusable for animated transformations such as stabilization. And it also depends heavily on the type of stabilization you're doing. If you do something like warp stabilizer it's very localized to different regions of the image so the crop is not an issue. BUT, you have no options as to what filtering is applied as typically warp algorithms don't allow for adjustable filter types. On the other hand, if you do a transformation based stabilization, it depends heavily on how much the camera moves at which point you may need a huge crop. But the best stabilization of them all is a cornerpin-stabilization, ala mocha or syntheyes 3d stabilization, and those require huge crops.
The technique that is typically used in VFX would be to degrain it with something like neat video, stabilize, sharpen and then add the grain back using either scanned grain or grain tools inside of your compositing application. The grain helps a lot to give the illusion of detail back into the image.