Found my usenet post about this from before (google good). Looks like I was
remembering how chroma sampling works incorrectly so I thought I'd post the
proper information from that post:
Well a big savings in the jpeg algorithm is the chroma sampling. That
creates a fixed reduction in size. If 2x2 sampling is used (apparantly,
Photoshop doesn't do this at higher quality settings, so it won't be an
issue with Photoshop), then you get a savings of 50%. This is achieved
by converting the RGB data into YCC data (luminance, Blue Chrominance,
and Red Chrominance). It turns out that the luminance value is the most
important part of this data. So the algorithm averages the other 2
values for each 2x2 block of pixels. So, instead of taking up 12 bytes
(4 pixels * 3 bytes per pixel) for each 2x2 block, you use only 6 bytes
(1 byte for luminance per pixel * 4 pixels + 1 byte for blue chrominance
+ 1 byte for red chrominance). This part of the algorithm will change
the colour slightly (really only noticeable if examined closely in an
image with quick changes in colour) but will not introduce any artifacts.
Andrew "frugal" Dacey
frugal@xxxxxxxxxxxxxxx
http://www.tildefrugal.net/
< This message was delivered via the Olympus Mailing List >
< For questions, mailto:owner-olympus@xxxxxxxxxxxxxxx >
< Web Page: http://Zuiko.sls.bc.ca/swright/olympuslist.html >
|