>> Every time you change the resolution, the quality gets worse, scaling
>> is a lossy process (if not scaled by exactly 2,4,1/2,1/4 etc.).
> That's not inherently true. Simpler scalers, which do not interpolate,
> should be able to upscale and then after downscaling offer very close to
> original video. sincR supposedly comes closest.
Still if you don't need it, I really wouldn't use it. It's comparable to
an A/D and then D/A conversion cycle.
In case of a movie that was shot with a sensor of a certain resolution,
you'd really want to have that dot-for-dot on your digital screen if
possible. On a CRT it would matter less.
> It's a rare, exceptional DVD that doesn't have noisy edges transitioning
> to a few pixels of black around the fringes, which is almost as bad
> as non-16 dimentions (less likely to triger codec bugs--border_mask
> may help).
In theory, yes. In practice I've actually seen more or less significant
changes in file size (yes, with same quality of course). This may be a
matter of taste.
> That is the recomended method of course, which is why cropdetect by
> default rounds down to multiples of 16. It can be problematic in a few
> situations, however, where subtitles or the like are right at the edge
> of the screen. Changing the offset (cropping more pixels at the top
> than the bottom) usually works, but not always.
Letterboxed material with wronly placed subtiles are the biggest pain of
all :-( (even on DVD's this common practice..)
No comments:
Post a Comment