So far as I can tell MHTML is HTML with the images and other files referenced by using relative paths instead of the canonical (absolute) path. For example if someone sent you an email page the path name to an image might be /images/myimage.gif and not http://mysite/myimage.gif. The images are encapsulated in MHTML files.
Scripting is also disabled in MHTML files.
There is an example of this at http://www.mail-archive.com/perl-win32-users@listserv.activestate.com/msg21313.html
Decoding and getting the images back out of the MHTML file may well be a problem. There is a program called SiteScooper that looks as if it may do the job you want. Although it was designed for Palm machines the home site also says it will run on PCs. See http://zap.to/sitescooper/