son of X-Face:

James Ashton (jaa101@syseng.anu.edu.au)
Tue, 17 Aug 93 9:43:52 EST

With `multimedia' computing taking off even for some pretty low end
machines it looks like X-Face is gaining wider acceptance. As the
author of the compface routines which define the `X-Face' standard I've
had contacts from people interested in using X-Face in products and in
extending the standard. Apparently customers want colour.

I'm happy to be involved in doing the work required to create an
extension to the standard. I can probably clean up the existing
compface routines at the same time. What I need is feedback about what
people want and what is practical. Also if I proceed I'd like some
sample data to play with. I can scan in colour here but images from a
single site tend to be similar. I'd hate to spend time tuning my
compression and then find that data from elsewhere compressed poorly.
I expect 48x48x8+colour map will be the easiest for me to work with.
I'm sure I can convert from any of the popular standards.

A quick list of issues that have occured to me are:

Resolution. Almost certainly still 48x48.

Colour map usage. Per face maps or a fixed map for all faces?
Other possibilities? How many colour map entries can users
afford? Can we support hardware with very few available
colours?

Should the compression be lossless as now or can we afford
lossy compression with colour?

Since we have colours will we use dithering? Dithering
obviously helps alot with a limited number of available colours
but it makes compression harder.

Will the images look the same on different hardware or will
more available colours allow a better image.

I'm assuming existing standards (JPEG comes to mind) are
unsuitable for this purpose. Please let me know if not and
save me lots of work.

How many data can we stand in the headers? Given that `X-Face'
is usually three to four lines it's hard to see a colour
version using less than six lines.

Backward compatibility is impossible without choosing a new
header title, say `X-Image'. Should an old style, one bit
image be available from an `X-Image' along with a colour
version or should we just include both an `X-Image' and an
`X-Face'.

Speaking of standards, this is all unofficial at this stage. What do
people think of moving towards a `Proposed Standard' RFC or at least an
`Experimental' RFC.

TIA for any feedback. I may post this to an appropriate newsgroup
depending on the response from this list. Finally, if you know of a
good reference on image compression that may be useful let me know.
There's likely to be some recent work I'm unaware of.

--
James Ashton						  System Administrator
VK1XJA					     Department of Systems Engineering
Voice +61 6 249 0681	  Research School of Physical Sciences and Engineering
FAX   +61 6 249 2698				Australian National University
Email James.Ashton@anu.edu.au			   Canberra ACT 0200 Australia