Aesthetics-Diven Online Summarization to First-Person Tourism Videos

Nowadays video blog (vlog) has gradually become popular. In scenarios such as tourism, many vloggers use mobile terminals to record first-person videos, which have redundant information, complex scenes and continuous shots. Such videos bring inconvenience to the creation of vlogs. The technology of video summarization can extract the main information of original videos. However, these works usually rely on large computing resources, which are not applicable for this scene. This paper focuses on artistic video summaries of first-person perspective videos in tourism scenes, and at the same time takes the resource constraints of terminal devices into account. And we propose an aesthetics-driven online summarization method of the first-person tourism video. In the first stage, the model uses Canny operator and HoG features to quickly generate summaries online. And in the second stage, it further extracts video segments based on the aesthetics like color and symmetry of images. The results of experiments show that this method can effectively carry out video summarization and aesthetic processing, and the prototype based on the method was implemented with an APP.