[XviD-devel] SAD vs MSE for ME (warning! acronym overflow!)
Christoph Lampert
xvid-devel@xvid.org
Thu, 5 Sep 2002 13:55:27 +0200 (CEST)
On 5 Sep 2002, skal wrote:
> I found (sleep-) time to quickly write down what I
> was thinking of yesterday, about SAD vs. MSE.
> It's at:
>
> http://skal.planet-d.net/coding/interpolation.html
>
> Any opinion whether it can be useful or not?
Hm, the idea itself seems okay, but as you mention, it's not possible
for use yet. First, for 2D I guess you can't use twice the idea the idea
of 1D. You would possible have to solve a small system of linear
equations. Not much work, but more than one skalar product.
I would be interested in a really 2D test.
Second, the aspect of rounding still worries me. If you end up at
a position a=0.27 it's not clear whether a=0 or a=.5 is better.
However, what you describe could be suitable e.g. for GMC, where higher
fractional pixel resoltion is possible.
Christoph