I'm not sure whether or not this is too late or even worth your time, but it seems to me that perhaps you might be better to use Mathematica's built-in image selection tool? If you have your image in the processor, you can click on it and an array of options comes up. From there you can select the region you want, copy it and then run TextRecognize on it to get the text, from which you can then output to wherever you want. It might take a little longer than having Mathematica do it automatically, but from my experience using functions like ImageCorners is problematic when text is involved.

It would probably take less time than marking the red spots and then coding something to create bounding boxes, in my opinion at least. This might not be what you're looking for, but hope it helps!