Robust Handwriting Extraction and Lecture Video Summarization

Video tools developed today, teachers can record lecture videos and upload these lecture videos to e-learning system themselves. However, some students may only do not understand some fragments but they have to waste unnecessary time download entire video, and therefore video scene segmentation is relative importance. In addition, in traditional teaching model, students must listen and transcribe content on blackboard, when the lecture's write speed is too fast, students are very difficult to pay attention in class and easy to transcription errors. If we can record the lecture video and automatically extract content on the blackboard, students not only enable index videos easily, but also not easily transcribe the wrong notes. Hence, this paper presents an intelligent assistance system for lecture videos. We use K-mean Segmentation to extract blackboard area and then only update this area to avoid lecturer's body cover the content on blackboard. Then we use adaptive threshold to extract chalk and design a method to reduce noise. After that we detecting the lecture videos split timing with statistics chalk pixel count and analysis the variation. Finally, we further interception the most complete content as lecture note to facilitate the students to search for video clips.