Re: Image compression with wavelets, DCT & spectral density

[Søren Hauberg]

ons, 26 03 2008 kl. 23:39 +0100, skrev MysticEyeBlink:
> I made two scripts for image compression with wavelet (WaveLab850) and
> DCT.
> My problem is: how to compute a spectral density and compression
> error.
I don't know what you mean by "spectral density", but when it comes to
compression the error, then the answer is simple: "you can't compute
it". This is the basic problem of image compression. We simply don't
have an error measure to optimise. If you look at the litterature,
you'll find tons of different error measures that you can compute.
Usually they are some variation of
  sumsq (im_orig(:) - im_compressed(:))
but they are all different, yet none of them are particular useful.

> The second problem is: my script to wavelet compression is too simply
> (example from wiki wavelab:
> http://wiki.octave.org/wiki.pl?WavelabOnOctave) and I don't know how
> to control compression factor. 
This is an algorithmic problem. You should ask people that work with
compression, and not on the Octave list.

> Please help :)
I think you are asking the wrong people. I haven't worked with
compression on a professional , so I don't know what forums exists.

Søren