Text Region Extraction Algorithm with Scene Selection for Automatic Text Translation Device

摘要

The needs of the wearable translation device is increasing for the international travelers. In the process of the automatic translation from the captured image by the wearable camera, it is important to extract the text region from the captured image as the first process. We use OpenCV image preprocessing library, modified stroke width transformation algorithm and connected-component labeling algorithm to detect text in natural scenes automatically. In this paper we are going to introduce the flow and the principle of this algorithm. Besides, we will discuss our idea about the automatic text translation device.