Our ~1500-page documentation set contains numerous screen shots and related graphics (schematics, flow diagrams, etc). Sometimes the user interface changes and we have to update all the affected graphics. The affected graphics are not necesarily all in one book; they can be spread across several. In short, they could be anywhere, so we rely on writers' knowledge of the doc set and sometimes just paging through the whole thing looking for graphics that are no longer correct. The first is fragile and the second is tedious (and also can be fragile).
Within the XML doc source itself I can (and do) embed internal tags that I can later search for. Think of that as meta-data for the docs. I'm looking for a way to associate meta-data with images, so I can find all the images that show such-and-such feature or such-and-such widget or whatever. I could set up an external index file or database, but that means the meta-data is far from the images and I worry about it staying up to data (will every writer always remember to update the database when he creates or edits an image and its meta-data changes?). Is there anything clever I can do to get the meta-data closer to the data?
Image formats are PNG for screen shots, GIF for line art. Is there any tool that would let us embed, and query, meta-data right into files in those formats?
Our existing relevant tools are: Perforce for source control, DocBook XML for the doc source, Ant for build targets. On the graphics side, we use PaintShop Pro and/or SnagIt for taking/editing screen shots, and mostly Visual Thought for line-art (though we have access to InkScape and can learn it if that would help). Our desktops are Windows (XP now, 7 later this year).
Edit: The screen shots also have text implications -- we usually don't just have a screen shot, but rather a screen shot and text that talks about the options or what you can do with that tool. The screen shots are integral to the documentation, and if a screen shot has to be updated we also have to look at the places where that image is used. Finding those points in the text is easy (I can just grep the source for the file name of each image when I update it), but finding the screen shots that have to change in the first place (because the UI changed) is much harder. This question is about managing that process so we can keep all the documentation up to date.