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

Peter J. Kootsookos wrote:
> "walala" <[EMAIL PROTECTED]> writes:
>
>
> > Why people think I want to lie upon seeing my question? Oh, it's my problem
> > that I did not clearly present the background...
>
>
> Partly.  Partly also you posed the question in a slightly troll-like
> way. :-)
>
>
> > Here is the story: in deblocking of block DCT coded JPEG images, it was
> > known that the DCTed coefficients are Laplacian distributed... But now I am
> > looking at low bit rate JPEG images, so there are someking of artifacts...
> > in order to reconstruct the original images... many algorithms have been
> > devised... one possibility is to make the image coefficients more Laplcian
> > like...
>
>
> These dudes appear to use the fact that the coefficients are Laplacian
> to a reasonable level of accuracy:
>
> http://bmrc.berkeley.edu/research/publications/1996/110/imdsp.ps
> http://citeseer.nj.nec.com/smoot96study.html
>
> So what makes you say your coefficients are not?  Are you looking at
> the right coefficients?
>
>
> > So that came my question: how to make data more Laplican like...Please give
> > me some detailed explanation as I am not veteran in statistics...
>
>
> Well, what variations do you have available?  Does your modification
> have to fit into JPEG? Or can it be seen as a completely separate
> step?
>
> Usually, if you want to convert a random variable with one
> distribution to another distribution you need to find a function that
> relates the two random variables.  
>
> A first, naive, approach might be to use the histogram of the data you
> have to find a piece-wise linear function that rescales the data so
> that its histogram is Laplacian.
>
> This technique is sometimes used in image processing to make greyscale
> images have a uniform brightness distribution (see e.g. 
>
> http://www.cs.tcd.ie/Fergal.Shevlin/courses/4d4/4BA10/PixelBrightnessTransformations.pdf
>
> )
>
> Ciao,
>
> Peter K.