In this paper, we found out a way to tell each character in movies efficiently. In addition, we cut the part that each character act in the movie.
In the mission, keeping to track the characters efficiently is the most important challenge.
When movies be played, we use Haar-like features to detect characters’ faces in the movie and capture them to classify. Every character’ face would trained by svm via EigenFace to complete the function of their faces recognition
We use Camshift Tracking to track characters in this paper. We use the feature of HSV ,tracking the middle position and weight to predict the object’s direction of movement. Moreover, we use projection to revise the middle position of the tracking scope to prevent from predict it wrong. After We Integrate identification system and tracking way, we put the characters in the tracking scope successfully through Face model. During the tracking process, we capture the part that each character act and get the part successfully.
Characters in the movies often pass each other, so it may cause mistakes when we track characters. Therefore, this paper used I-B-P, intra pictures, predicted pictures, bi-predictive pictures and bi-directional pictures through move vector to predict the direction of movement of the object that is covered. In addition, we move the tracking frame properly to revise the process when the objects are covered.
