From the OpenXML Developer site:
Robert Orleth, a development lead on the web content management team for Microsoft Office Sharepoint Server, has submitted an article on how to extract images from Open XML documents that reside in a Sharepoint document library and turn them into external linked images instead of internal images. This provides a clever workaround for the fact that the built-in convert-to-HTML functionality doesn't have support for images: since the images are external, they'll still work in the converted HTML. See "MOSS: Converting documents with embedded images to web pages" for all the details, including a walkthrough of the .NET code required.
Excellent stuff!