A Computer Vision based Whiteboard Capture System

Conventional whiteboard video capture using a static camera usually results in a poor quality. In this paper, we present an autonomous whiteboard scan and capture prototype system, which consist a pair of static and pan-tilt-zoom (PTZ) cameras. The PTZ camera is used to scan the newly-updated whiteboard regions without interrupting the instructor. We will illustrate several computer vision techniques used in our system: Firstly, we present our unique camera calibration method using rough hand-drawn gridlines. Secondly, we present the image processing methods used to determine where the newly updated whiteboard region to be scanned is. Our method also accounts for the whiteboard region occlusion from the instructor.