Currently, the scaled size of the image is ignored when parsing images out of a document. This causes the resulting HTML to have massive images and not match up with the style of the original document..
The (deprecated) python doc2html library did a nice job of this by resizing the images.