Well, not really. It won't perform ocr, it only works if you highlight actual text. Text on images won't be read. Actual text is already selectable, so this isn't all that useful.
You used to be able to smart select an image with an address, extract the text and have it input in maps for...