1 - Extract the metadata information from the meta.xml file for each image. You can do that by creating an XSL file that will be called from the ant task generated by xslt/get_ressources_urls.xsl. In that file you have the hand on each meta.xml File.
2 - Verify that images are well zipped to avoid any problem while testing in Opale
3 - Images inside paragraphs break the validation of the hdoc schema, do a preposition to change the schema and handle that.