If you want to do it totally yourself then you will need to read the RTF specification, particularly pictures and parse the contents;
Here is the spec for 1.5;
http://www.biblioscape.com/rtf15_spec.htm#Heading49[
^]
(Here is a link to the V1.9.1 spec at MS:
http://www.microsoft.com/downloads/details.aspx?familyid=dd422b8d-ff06-4207-b476-6b5396a18a2b&displaylang=en&tm[
^])
As you will see in the spec, picture are under the \pict control tag, and the tag and all the data are enclosed withinn { } brackets, so it is simple to then pick out the data required. From this you can parse out the sub tags to read the type of image and then reconstruct it by reading the bytes.
Sample rtf file data for a picture element;
{\pict\wmetafile8\picw9604\pich7064\picwgoal5445\pichgoal4005 <br />
010009000003c039020000009739020000000400000003010800050000000b0200000000050000<br />
000c020b016b01030000001e0004........data continues........ffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff<br />
ffffffffffffffffffffffffffff000000040000002701ffff030000000000<br />
}