@DBrant (WMF) and I have been discussing how images should be represented in the compilation info response. The Image type defined here might not make much sense:
1) Will the dimensions actually be useful?
2) It's looking likely that we'll bundle all of the needed info (including image/thumb bitmaps) directly into our versions of the ZIM files so that they can function as a self-contained unit. Would should the corresponding image/thumb treatment be in the compilation endpoint response?