Image-based electronic editions enable researchers to view and study in an electronic environment historical manuscript images intricately linked to edition, transcript, glossary and apparatus files. Building image-based electronic editions poses a two-fold challenge. For humanities scholars, it is important to be able to use image and text to successfully encode the desired features of the manuscripts. Computer Scientists must find mechanisms for representing markup in its association both with the images, text and other auxiliary files and for making the representation available for efficient querying. This paper addresses the architecture of one such solution, that uses efficient data structures to store image-based encodings in main memory and on disk.