I'd like to be able to highlight a word in an image of a document when the user searches for that word. Exactly like Google Books does here.
As far as I know, Tesseract and other open source OCR programs don't support this sort of function, so does anyone have any ideas how it might be done?
Yes they "support" it. Sort of.
They give you a rectangle that tells you where the word is. Using that, fill said rectangle with the color of your choice on the image using a color blending mode (e.g., keep the luma intact and just alter the chroma). This works well with B/W and grayscale images, which most books are, and is sufficient for most colored fonts too (except those in a colored background). A solution to this is to invert the colors instead of highlighting them, this is done in many applications (Foxit Reader comes to mind).