I suppose one option would be to build a script that opens the HTML in a web browser and then takes a screenshot of the web browser window. This is likely to be slow and non-portable, but I think it would work. If you could do it offline and cache the images it could probably be made to work.
A better solution might be to try to interface with an HTML rendering engine (like the Gecko engine used by Mozilla) and use it to generate a rendering in an image. This could be a very difficult task, but the results would likely be much better. If you go this route, consider building a Perl module that others can use to do the same task. A Perl interface to Gecko, for example, would be a popular module!
-sam
