A novel fused algorithm for human face tracking in video sequences
S Ranganatha, Y P Gowramma · 2016
In video processing, face detection and tracking has a wide scope for research. Although existing algorithms serve the purpose of face detection and tracking in video sequences, tremendous development in video technologies is posing more challenges while processing such videos. In this paper, we propose a new face tracking method based on fusion of corner measure algorithm and KLT tracker. Initially, Viola-Jones algorithm detect the face present in the first frame of the video sequence, detected portion of the face is extracted and Harris corner measure algorithm applied. Once corner points are computed, centroid is calculated. The resulting corner points which are in the form of an array are converted into matrix. KLT is a point tracking algorithm, works based on Eigen values and its result is in the form of matrix. The matrix generated from corner measure and centroid computation is concatenated with the matrix obtained using KLT face tracking algorithm. If the corner measure algorithm fails to track the facial region then KLT tracker performs its work and the reverse is also true. The proposed algorithm achieves accuracy better than KLT algorithm alone. The results at the end of this paper clearly indicate the practical achievement of theoretical assumption made.