Text this: Extraction of Key-Frames From Endoscopic Videos by Using Depth Information