Improved Time-Saving Face Detection in Video

Human face detection from video sequences is a difficult problem in computer vision. Face detection is the process of determining the location of a face or faces in each frame of a video. Face detection in real-time videos is an ongoing under-investigated research problem. This paper proposes a new method for face detection in videos that significantly reduces the time required to detect faces. The method targets a Search Area Boundary (SAB) in the next frames that are determined based on the position of the detected faces in the current frame. The model will look for the face in a box that is slightly larger than the SAB from the previous frame. This means that it will not search the entire video frame for the face, so it will take less time to find the face. After a few frames, the method will begin searching the entire frame for new faces. The results show that the proposed method can save up to 73.5% of the time, especially when the SAB is small, with an accuracy degradation of no more than 7% in the best case.