Wavelet decomposition of images

I'm wondering if an image such as http://en.wikipedia.org/wiki/File:Jpeg2000_2-level_wavelet_transform-lichtenstein.png is just a representation of a wavelet decomposition, or if you could actually recover the decomposed image.

The decomposition code works with doubles, so while I can decompose an image properly (each colour channel separately), the resultant values are doubles and some of them are outside the 0-255 range. I'm not sure if it's possible to represent the result as an image and then reverse the decomposition, since I'd have to transform the values to bytes and I'd lose a shitload of precision.

this is now a post ending in doubles must post tits thread

>>1
We don't do doubles on /prog/, polecat keba/b/s.

>>1
Maybe you should invent discrete-discrete image processing and work with that.

I don't think you would lose that much information. 0-255 all that the computer can display anyway. So it doesn't matter.

<--- check 'em flotz

<--- check em' intz

>>3
Fuck off and die, ``fagstorm''.

>>4
I'm trying to implement a denoising filter and I think that I can simply work entirely with doubles; I'm just wondering if images such as the one linked in the OP are only representations of the wavelet decomposition or if they're actually usable.

>>8
You should be working with doubles because that's what the wavelet transform returns.

For fourier transforms atleast, you can store the resultant frequency image as an actual image. And I don't think you lose all that much information after you round all the numbers (ofcourse you do lose information). The inverse transform would work fine, with a bit of noise added. Same goes for wavelet transforms, I suppose.