The actual standard allows for a variety of options around how the compression portion is handled -- there is the baseline mode and then options for custom huffman tables, etc. The best source for the JPEG Standard --
My memory is a bit fuzzy here, but from what I recall, a friend tried implementing JPEG compression on a parallel system (back in the mid 90s). One of the trouble spots was the reliance on the relationship between blocks regarding prediction of the DC coefficient.
One startup I worked at, we had an entropy encoding method that could add 20-30% additional compression just by recording coefficients. Patent here -- http://www.google.com/patents/US6236762
http://www.amazon.com/JPEG-Compression-Standard-Multimedia-S...
My memory is a bit fuzzy here, but from what I recall, a friend tried implementing JPEG compression on a parallel system (back in the mid 90s). One of the trouble spots was the reliance on the relationship between blocks regarding prediction of the DC coefficient.
One startup I worked at, we had an entropy encoding method that could add 20-30% additional compression just by recording coefficients. Patent here -- http://www.google.com/patents/US6236762