the images are seperated by channel. man image(6) and draw(3).
a 24-bit image is 1 byte each r, g, b premultiplied by the alpha channel
which is in the fourth byte.
but they are not contiguous in memory, and that is the "usual" way to
split an RGB for when you want to manipulate them
surely you'd want to undo the premultiply of the alpha
What are you going to do when you want to move the RGB into HSB or CMYK
space ?
The best approach, in my opinion (and it's only that), is to spearate
them into channels and then you can have a set of operations that works
on combining channels and doesn't care about what that channel is
representing.
A bitmap of 8 bit values represents R,G,B,A,H,S,B,C,M,Y and K and any
others you come across.
Having them interleaved as RGBA, RGBA, RGBA will give you a headache
sooner or later.
Split them up, get them away from a particular image format be it plan9
bitmap, tga or whatever and then you can choose to write out / read in
whatever formats you like.
Just my experience of working in broadcast video processing, YMMV.
Matt
