Enhancing Readability of Scanned Picture Books

I describe a system that enhances the readability of scanned picture books. Motivated by our website of children's books in the International Children's Digital Library, the system separates textual from visual content which decreases the size of the image files (since their quality can be lower) while increasing the quality of the text by displaying it as computer-generated text instead of an image. This text-background separation combines image processing and human validation in an efficient manner and results in a system that not only is more readable, but also accessible, searchable, and translatable.