Re: Indeo 4.x and 5.x

Hi.

Jim Leonard wrote:

> I was always under the impression that Indeo 4.x and 5.x were 
> wavelet-based -- I was wrong!  They use a block slant transform, 
> similar (if not identical) to a Haar transform.  From a Usenet posting:
>
If you search the archives for Indeo information you will find some more 
info about IV4 and IV5.

> AFAIK Indeo 4.x uses 8x8 block slant transform.
> It is less optimal for compression then DCT, but can be implemented
> without multiplications - big advantage for an old 486 CPU (16*16bit
> imul takes up to 26 cycles on i486, shift - only 2 cycles).
> Apparently, slant transform is the most efficient one from the
> transforms not based on sinusoids.
> So, Indeo 4.x is pretty much the regular block transform + motion
> compensation kind of codec. It's just the transform is sort of
> "wavelet-like".

Correct, but the basis looks quite like the Descrete DCT.

Page 8, 
http://www.icg.isy.liu.se/courses/tsbk01/material/slides/TransformCoding_ee392c_Girod.pdf

>
> Does anyone have any additional information on "block slant" 
> transforms?  I would love to learn more, but all I've found on the web 
> is http://rkb.home.cern.ch/rkb/AN16pp/node113.html#112

Just search for "Slant transform image coding" on google.

MvH
Benjamin



-------------------------------------------------------
This SF.Net email is sponsored by: GNOME Foundation
Hackers Unite!  GUADEC: The world's #1 Open Source Desktop Event.
GNOME Users and Developers European Conference, 28-30th June in Norway
http://2004/guadec.org