Face tracking with subtracting frames in video

Question

Is it possible to track a face in video with subtracting frames without using face recognition?

What happen if the face change in next frame? Is there any way to detect this change with subtracting?

Answer 1

Face tracking is different than, face recognition. Simply,

Face tracking means, tracking an object that has features of a face.

Face recognition means detecting and recognizing a face among a set of already known faces.

For tracking a face firstly, you need to detect it. So, for detecting a face there are simple techniques such as Haar Feature-based Cascade Classifiers, and LBP cascade classifier. You can google them and read about them.

After the face is detected, then you can try to solve the problem of face tracking. But tracking the face through different frames, means that you repeat the face detection process for each frame. Now the question would be how to increase the speed of the detection, that be suitable for a normal frame rate like 30 FPS? A simple solution is to decrease the search area. In other words, if the face is detected in the first frame, in the second frame there is no need to search the whole area of the frame. The optimum solution would be to start the search from the position of the face in the previous frame.

A simple face detection and tracking tutorial can be found here.

Answer 2

Try this example, which uses the Viola-Jones face detection algorithm, and the KLT (Kanade-Lucas-Tomasi) algorithm for tracking.