Question about importing data from memory

Hi,

I am processing some imagery that comes directly from sensors so that I 
end up with it in an NxM array of unsigned integers of rgb values (for 
example, array[0][0]=\xff0000 would be a red pixel in the upper-left 
corner). Some of this imagery wants to end up on web pages with text 
annotation, which is why I am looking at using the gd library. 
Currently, I save these images as bmp files using my own code, convert 
them to jpeg using imagemagick's convert utility, and they have no 
annotations.

Is there a way to get the data from my arrays directly into a gdImage? I 
was looking through the API and could only find the function 
gdImageCreateFromGdPtr:

http://www.libgd.org/OldImageCreation#gdImageCreateFromGdPtr.28int_size.2C_void_.2Adata.29_.28FUNCTION.29

but I couldn't find any description of the format of a GdPtr. I could go 
to the source code, but in my experience when something isn't described 
in the API, its often not stable, so I don't want to get off on the 
wrong track.

I did take a look at a couple of the files like gdxpm.c for guidance, 
and there it appears that one creates a gdImage and populates it by 
creating a color table on the fly and writing indices instead of pixels. 
The problems I can see with this approach are that - unless duplicate 
color entries are allowed - there will be a significant overhead 
involved in searching the table for an existing color each time before 
creating a new one, which with Truecolor could be hugely inefficient.

 I didn't see any related discussions in the list archives. How do 
people handle this kind of problem?

Thanks,

Jeff Cunningham