JPEG Variable Quantization
Pete Fraser
the quantizing matrix variably throughout the image.
Did this ever take off?
I've found a variety of papers, but does anybody use it?
Matt Mahoney
It's not part of the original JPEG spec AFAIK. How would you implement
it and be compatible?
-- Matt Mahoney
Pete Fraser
"Matt Mahoney" <matma...@yahoo.com> wrote in message
news:1119574435.9...@g43g2000cwa.googlegroups.com...
The choice of scale-factor table is done with a new marker segment
(DQS) that would presumably be ignored by an unaware decoder.
The signaling of a quantizer scale change is done by adding
a new symbol to the DC table (QS_CHANGE = 0x15).
If QS_CHANGE is present, the following 5 bits give the index
into the table of Q-SCALE vs. SCALE_CODE selected by
the DQS marker segment. I assume an unaware decoder would
barf on the extra signalling.
None of the small number of codecs I've looked at supports variable
quantization. I'm looking at this for warped fun, rather than serious
use, and thought somebody might have used it somewhere.
[It must be horrid generating a standard that nobody bothers with.]