Thread Subject: 2-D DCT

Hello everybody,

Fast Matlab 2-D Discrete Cosine Transform released at:

http://www.mathworks.com/matlabcentral/fileexchange/loadFile.do?objectId=21656&objectType=FILE

Regards,
Bodorin

jaja...

Ok, but:

1. there is not h1 line
2. there is not description about what is this for
3. it is not vectorized
but, most important
4. not uses fft, so... as it is, not worth it.

Carlos

Nothing to say ... except this:

http://www.cs.cf.ac.uk/Dave/Multimedia/node230.html

Bodorin

"Carlos Adrian Vargas Aguilera" <nubeobscura@hotmail.com> wrote in message <gc3cd9$5pr$1@fred.mathworks.com>...
> jaja...
>
> Ok, but:
>
> 1. there is not h1 line
> 2. there is not description about what is this for
> 3. it is not vectorized
> but, most important
> 4. not uses fft, so... as it is, not worth it.
>
> Carlos

Hello, Carlos,

An m-script comparing FFT-based DCT2 with BLAS-based DCT2 was released here:

http://www.mathworks.com/matlabcentral/fileexchange/loadFile.do?objectId=21822&objectType=FILE

meaning that ...
most important,
FFTW is the best fft library I now so far but it isn't some kind of universal solution for any fft-related computation.

DCT2 in itself is much simpler than the fft and BLAS library is still alive ... providing very powerfull tools for matrix operations in Matlab.

By taking the above link and running the code, you will see why at least "block_idct" function couldn't be outperformed by some fft-based m-script computation of IDCT2.

Regards,
Bodorin

P.S.
Until testing, don't get tired judgeing so hard!
isn't worth it ...



"Carlos Adrian Vargas Aguilera" <nubeobscura@hotmail.com> wrote in message <gc3cd9$5pr$1@fred.mathworks.com>...
...
> but, most important
> 4. not uses fft, so... as it is, not worth it.
>
> Carlos