[libdv-dev] Re: DV codec: quality problem

[Ardi <ago...@wa...>]

Thank you for your answer. Also to Dan Dennedy for interesting links that he
sends to me.
I've made some complemantary tests that confirm my first analyse and also
shows that hardware decompressors works better, for this issue, than
existing software ones. I've updated my web page in this way.

I will be very interested by your results if ever you run my test with your
software, and try to understand the origin of the trouble.

Is it possible to find somwhere (preferably on the web) the process of DV
compression? (in theoritical point of view).

Best regards
Ardi

----- Original Message -----
From: "Charles 'Buck' Krasic" <kr...@ac...>
To: "Ardi" <ago...@wa...>
Cc: <kr...@ac...>; <lib...@li...>
Sent: Thursday, January 31, 2002 10:40 PM
Subject: Re: DV codec: quality problem


>
> Hi.
>
> In my experience writing libdv, it was very easy to have bugs which
> lead to artifacts such as you describe, and very difficult to track
> them down.  I could hazard a guess at the source of the artifacts you
> see: either they are not handling 2-4-8 dct blocks correctly, or they
> are not handling class 3 quantization correctly.  But that is only a
> guess.  Tracking these kinds of bugs down is very tricky.
>
> If I had time, it would be interesting to port libdv to be a Premiere
> plugin and compare the results.  I am confident that our quality is
> very good.  Of course, I am also afraid that our speed is worse than
> the commercial codecs.
>
> Cheers,
>
> -- Buck
>
> ps You have gone to a lot of trouble to clearly identify the problem.
>    A nice thing about open source is that someone with enough
>    motivation and skill can go all the way to try and fix the bug.
>    Not an option with Pinnacle et al.  :-|
>
> "Ardi" <ago...@wa...> writes:
>
> > Hi,
> >
> > We have here an amazing problem with DV codec.
> >
> > Until now, all Pinncale, Microsoft, Main Concept and Panasonic codec we
have tested produce very poor results.
> >
> > I explain this in the following page:
> >
> > [[http://golgolabprod.multimania.com/CODEC-problem.html]]
> >
> >
> >
> > Do agree my analyse or have an idea about this problem?
> >
> >
> >
> > Bye
> >
> > Ardi