Re: AW: annoying E effects on horizontally moving items

[Matthias Oelmann <moelmann@ixl.com>]

Schuett Thomas wrote:
> 
> Hi Matthias,
> 
>       >I'm currently working on that problem - the code is not finished yet
>         >though.
> 
> That is really great!
> But the algorithms you mention sound worse than my proposal.

hmmm...

> 
>         >The easiest way is to do an interpolation between the lines of
>         >one field (halfpicture) to get twice as many lines (linedoubling).
> 
> So you mean, you just do not use the other half picture. This throws away
> a lot of information.

No, you dont throw away any information.

Imagine a tv picture A consisting of 576 lines. the first field (A1)
would consist
of all even lines ( 0, 2, 4, ... ) , A2 would consist of all odd lines (
1,3,5 ...).

now take only A1 - it consists of 576/2=288 lines. scale this by factor
2 (vertically) 
to get 576 lines again (line doubling) then display it for 1/50th
second.

take A2, do the same, and display it with an offset of 1 line (
important ).

you will get some artifacts ( flickering ) on sharp edges, but it 
is simple to do and looks a lot better than before.

> 
>         >If you want to do it right, you weave those parts of the image where
>         >the is no motion and interpolate the parts with motion...
> 
> This would be most perfect, but really CPU intensive and hard to code.
> 
My goal is to get the very best quality possible - if you use hardware
mpeg decoding you will have enough CPU cycles for this.



> Please think again about my idea:
> Imagine, where the interpolation of b-picture[t] with b-picture[t-2/50]
> would be. It would be similar to a-picture[t-1/50], would'nt it?
> And the full information would be available on the screen.
> And think again about the effect on still pictures. It would be "perfect",
> I mean, it would be "as unchanged".
> 
it is an interesting idea and i might try it out as soon as i have my 
image processor framework working but i think it will give you bad
artifacts.

think about a black box flying over a white surface
(X=black pixel, .=grey pixel)


picture 1a
------------------------------- < top of screen
      XXXXXXX

      XXXXXXX

      XXXXXXX
 

picture 1b
------------------------------- < top of screen

         XXXXXXX

         XXXXXXX


 
picture 2a
------------------------------- < top of screen
            XXXXXXX

            XXXXXXX

            XXXXXXX


picture 2b
------------------------------- < top of screen

               XXXXXXX

               XXXXXXX



your algorithm for 2a
(interpolation of 1a & 2a, weaved with 1b)


------------------------------- < top of screen
      ......X......
         XXXXXXX
      ......X......
         XXXXXXX
      ......X......


see.
but really, this is only theory - it might look better 
than this. or did i understand something wrong?

what graphics card do you have? i do have a driver almost working that
implements the method i explained above in hardware on
a matrox g200/400 - if you are interested...

have fun,

Matthias


> Maybe I can win you for this, or you can tell me bad side effects of
> my algorithm.
> 
> cu,
>   Thomas