Home > Archive > Compression > January 2007 > Re: issues, DCT vs. MDCT (Re: misc, possible promise: MDCT and image
You are viewing an archived Text-only version of the thread.
To view this thread in it's original format and/or if you want to reply to
this thread please [click here]
| Author |
Re: issues, DCT vs. MDCT (Re: misc, possible promise: MDCT and image
|
|
|
| cr88192 ha scritto:
> "Carsten Neubauer" <cn@c14sw.de> wrote in message
> news:1168561573.702996.294540@a75g2000cwd.googlegroups.com...
>
>
> odd, I was getting by fine having an all the pixels as bytes...
>
> then again, I am facing other issues here, most notably the reduced quality
> at the higher end (and when zoomed in, high-frequency noise).
>
> it is possible though that this is a result of doing my various transforms
> using integer math (and not only that, not very precisely, trying to avoid
> potentially overflowing an int...).
>
>  ly, due to range limitations I ended up having to do an extra set of
> shifts.
>
>
> as implemented (approximate):
>
> IMDCT total cost is somewhere around 8 multiplies, 8 adds, and 2 shifts per
> pixel (excluding YUV->RGB conversion, ...).
>
> IDCT implies 5.5 multiplies, 9 adds, and 1 shift per pixel (as above).
>
Have you tried doing this stuff on a GPU instead ? :)
Best,
E.
|
|
|
|
|