In order to incarnate “multithreading” of news event reporting, the paper studies the validity of graph theory for managing news video database. Firstly, news stories similarity and graph theory are studied for digesting complicated relations of all stories reporting the same event. Then, a method based on minimal spanning tree is presented for predigesting the similarity graph into a simple tree with “multithreads”. Experiments show that the method is a research issue with great theoretical and pratical significance for browsing, searching and abstracting in news video database.