I did this with OCR text. The OCR software output bounding boxes which I converted into client side image maps. The company wanted to patent it.