Hi,
I was a games programmer from the mid 80s to the early 00s so I freely admit to being 'out of the loop'. One technique used in the Tomb Raider 2 and all the following Core Design Tomb Raiders was originally designed to run on a board boasting 2 x 200MHz StrongArm processors for real-time encoding and decoding. The guys had also produced an 'analysis by synthesis' methodology so a frame wasn't compressed down to a maximum size but rather up to a minimum quality. That meant that later versions of the compression codec took about 45 minutes on a top-end PC (a 200 MHz Pentium) to compress a 3 minute video.
I've noted that in 2003 onwards, many researchers have implemented versions of analysis-by-synthesis (and I would love to read all of the papers but they cost >$30 each and you don't know which ones are of use UNTIL you buy them. I guess someone doing postgrad work would be able to access them all free as different departments in universities pay for access to all of the back issues of various journals.
Has anyone looked into a codec in which the image is encoded UP to a quality rather than down to fit a size? My interest is first to check that nobody else has done this and I'm wasting my time and to garner opinion of those with experience of this format.
Thanks for your time,
MC