Moose wrote:
> I have no idea how JPEG compression works inside.
See http://en.wikipedia.org/wiki/Jpeg#Encoding for a (for wikipedia)
surprisingly good explanation, especially if you ignore the giant
matrices in there and just look at the pictures.
Basically, it splits the image into luminance, chrominance-red,
chrominance-blue (same encoding as PAL). (YCbCr)
Then it (possibly) throws away some detail in the Cb/Cr areas because
the eye can't percieve as much detail in changes of colour as in brightness.
Then it splits the image into 8x8 tiles (which you can see if you save
a jpeg a really low quality), and does a DCT on the resulting tiles to
turn the info from pixels into frequency info. (the same sort of thing
as generating the fourier series for the image, if that's any more
meaningful).
Then, depending on how much compression you ask for, it throws away
high frequency info. (this is why jpeg loses sharp edges between bits of
image, because to encode that, you need high frequency info -- but for
photographic images, there are normally no such very sharp transitions).
Then it compresses the resulting data using a lossless compression
method; this doesn't affect the image any more.
-- dan
==============================================
List usage info: http://www.zuikoholic.com
List nannies: olympusadmin@xxxxxxxxxx
==============================================
|