Re: (statistics)how to make date more like Laplacian distribution?

Dear Marco,

> They have compared the transform to DCT with a drop in replacement in
> standard JPEG ... given what you said I doubt your DCT is standard
compliant
> though, so that still doesnt allow you to make a real comparison with your
> own transform.
>

I just wanted to say that binDCT itself does not give good quality, it is
the EZW-like packetization scheme make it good PSNR... If binDCT is then
quantized with normal JPEG quantization matrix, that will give lower PSNR...
I guess...

>
> > function y=mybias(x, Q)
> > sigma=abs(x.*Q);
>
> That is a strange way of calculating variance ... what exactly are you
> trying to accomplish there?

No I am using the exact formula in that paper to compute variance... it says
that variance = (F_reconstructed)^2... since F_reconstructed is the
coefficients after dequantization, and my x, y are all for quantized domain,
so variance=(x dequantized )^2, right? Then sigma=abs(x dequantized)...

>
> Dont forget BTW ... DC does not have a zero centered laplacian
distribution.
>

Yeah, good point! I forgot to exclude DC in my bias calculation... I should
set that mybias(0,0)=0 directly...

Thanks a lot,

-Walala