[MPlayer-users] Re: RFC: docs update for "how to create a highquality DVD rip"

[D Richard Felker III]

On Fri, Jun 11, 2004 at 02:45:44AM +0300, Ivan Kalvachev wrote:
> D Richard Felker III said:
> > On Thu, Jun 10, 2004 at 03:36:33PM +0200, Tobias Diedrich wrote:
> >> D Richard Felker III wrote:
> >> > Yeah, it's actually amazing how much DVDs suck. BTW for those complaining
> about blocking artifacts and such in their lavc mpeg4 rips, guess what?
> The original DVD has blocking too... :(
> >> >
> >> > The solution? Rent the film cans and use professional equipment to
> capture it at hdtv res (or higher) and with 16 bit luma precision, then
> make a new codec that allows 16bit luma... Or just be happy with what
> you've got.
> >> AFAIK mpeg2 does allow up to 10bit dc precision, but with current
> hardware/software you are limited to 8bit on input and output... However I
> don't know how big the difference between 8 and 10 bit really is be.
> >
> > Well the blocking problem seems to be that with (stupid!) video
> > hardware (with improper gamma!) you can visibly see the difference between
> consecutive intensity levels at 8bit precision. As long as this is true,
> you're doomed to have some sort of artifact, whether it's blocking or
> banding in slow gradients or whatever... A few more bits would help a lot,
> but like you say everything is limited to 8.
> >
> > Rich

Could you fix your mailer? It's wrapping quotes totally wrong. See
above. It shouldn't rewrap them at all, but leave them as-is... Makes
it very hard to read...

> It look like you are messing something.
> The 10bit DC precission is on DC component of the IDCT coeficients, And all
> coefficients are 12 bits (max). As DC is coded as difference, it is not always
> bad issue (e.g. it is like using quantizer 4 at worst case)

Quantizer 4 will look absolutely horrible on a picture with
almost-contant intensity...

> About scanning film. Yes they scan it at monster resolutions, but this is
> nothing compared with postprocessing (an probably hand work - look at
> filmgimp)
> to remove the film grain and to sharpen the edges.
> 
> And additional bit depth is used for fixing and brightness/contrast issues.
> That's why scanning is always done with more than 8 bits per channel.

Yes...

Rich