non_max_suppression overlapping boxes

Question

I am using a tesseract for text detection. The algorithm gives as output many boxes, and often overlapping. I am using the non_max_suppression from the imutils library.The problem that I have is that the function compares the surrounding boxes, and picks one result. What I would like is that the algorithm joins all surrounding boxes and return one boxes that covers all the previous one. Does anyone knows an algorithm to achieve this?