In video indexing and retrieval systems, textual information is high-level semantic content to describe images or video frames, which is also one of the most interesting objects of study. A great progress has been made in text detection in recent decades, but it is still a great challenge due to the complex background, various fonts and different languages, etc. In this paper, we propose a edge-based approach to detect artificial text. First, the image edge map and difference range map are calculated. Then the map is binarized by local threshold method. Finally the textual information is located through the regional projection analysis. Experience results show our method’s precision rate is 91.1% and recall rate 93.2%, which outperforms the previous work.
ResNet CNN with LSTM Based Tamil Text Detection from Video Frames
