Downsizing 1080p>720p what color sampling 4:2:2?

M

Moses Chisel

Active member
I am experimenting with downsizing 5D2 footage from 1080p to 720p. If I apply some circular blur and expand the color space (carbon coder) before downsizing, the footage looks quite pretty and smooth in Edius using HQ codec.

I was wondering what color sample scheme does the 720p become. Is it 4:2:2 since the image area is approximately half?

Also I note some are saying the 1080p is really embedded 720p anyway, so do I lose resolution by downsampling?
 
Moses, you're asking a question that it's not that easy to answer, but I'll take a stab at it...

The footage the SLRs produce is about 720p of resolved detail, embedded in a 1080p frame, with a whole lot of spurious aliasing added on top. When you resize it down to 720p, it doesn't just chop off the aliasing and leave you with a clean 720p, the aliasing gets factored in. Once it's in there, it can't be gotten rid of. So you'll have a good 720p with also aliasing folded in on top.

As for color sampling -- the color sampling will actually look more accurate at the resized 720p size, versus the original 1080, because the color matrix doesn't have to lose detail during the resize. 4:2:0 1080p has a color matrix of 960x540, so there's actually more than enough resolution in a 720p frame to hold all the color detail, but there's not enough color detail for the whole frame. So it won't be 4:2:0 or 4:4:4 or anywhere really close to it. But it'll be better than it was when it was spread across the 1080p frame. But it's hard to quantify.

In the source 4:2:0, one color sample has to average together 4 pixels, in a 2x2 matrix. But in resized-down-to-720p, each color sample would technically have to cover 2.25 pixels, in a 1.5x1.5 matrix. But of course there aren't half-pixels, so that doesn't really count, so ... I don't know how to quantify it for you, other than to say that the color sampling will be somewhat more accurate on a per-pixel basis in the resized footage, versus the original.

(and this talk of "color matrix" isn't meant to be technically accurate, only for visual illustration purposes).
 
Hi Barry,

I am finding the expand color space and circular blur do seem to combine to give an overall better image (imo) after downsizing. I realise the bayer thing in the CMOS doesnt really give 4:2:0 in the first place.

One thing I note is the aliasing is significantly reduced but not illiminated.

best,

Mo'
 
You must log in or register to reply here.
Back
Top