JPEG press release

http://www.jpeg.org/newsrel1.htm

A similar list can be put together for video tools.

You will be very hard pressed to do better than H.26L
on a modern SIMD CPU.

Wavelets, fractals, object coding have been bandied
about for more than 10 years now, but have not made a
dent. JPEG-2000's entropy coding (coefficient packing
& arithmetic) provide the small gain over JPEG-1 ---
not the wavelet transform itself. The subbands provide
a framework for the author to play games with
subnyquist "sampling" that would otherwise be
performed with preprocessing and proper content-sample
rate balancing with JPEG-1.

These alternate tools do provide a different
perspective that allows the correlations and psycho
visual modes to be expressed differently. However, if
you focus at the low level operations, they do not
provide much over state-of-the-art block coding
methods. Block coding simply sees a picture as
collection of subpictures, each potentially with their
own statistics. Yes, blocks cross object boundaries
(if the block shape is not sufficiently adaptive), but
it takes overhead bits to prescribe the outline of an
object, and the object deforms over time .. more
overhead.

For example, a struggling well-known object video
coding company that claims 4:1 better compression over
MPEG-2 (which in reality is only 2:1 when all factors
are taken into account) requires hardware with 30x the
complexity of MPEG-2. H.26L is genuinely 2:1 over
MPEG-2, but only 5x the gate complexity of MPEG-2.

You get what you pay for, as always. A safer path may
be to find what works in well known coders, and
substitute the patented parts with prior art, or other
work-arounds.



__________________________________________________
Do You Yahoo!?
Yahoo! Health - Feel better, live better
http://health.yahoo.com

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to '[email protected]'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.