Pik is to have roughly the same performance requirements space that
JPEG is holding today. We aim to have roughly similar decoding speed,
i.e., no more than 50% slower in decoding, but adding more modern
compression and prediction methods, giving 55% more density. We plan
to improve to 65% before freezing the format. Some of these new
advances may come with a moderate drop in decoding speed (possibly 40%
of jpeg speed).
We are planning for a tiled format so that multi-core decoding can
decode a single image faster (possibly around 16x). The current
version has no support for this yet.
We are planning to keep the format 8x8 DCT based, possibly with some
support for non-integral-transform-based direct mode blocks (or overlay