Shot boundary-based keyframe extraction for video summarisation
G. G. Lakshmi Priya, S. Domnic · International Journal of Computational Intelligence Studies · 2014
We propose a new automatic video summarisation technique using shot boundary-based keyframe extraction method. Initially, the frames are segmented into shots by identifying the transitions using a new CEM-visual content descriptor (colour, edge strength, moments). From the segmented shots, motionless visual content representative frames known as keyframes are extracted. However, the number of keyframes to be extracted for each shot is determined using the shot dispersion ratio. The proposed shot boundary detection algorithm and video keyframe extraction technique are implemented and evaluated over TRECVID dataset. Compared with existing related algorithms, our algorithm yields better Fl-score of 94% for shot boundary detection. Also, our algorithm yields average fidelity value of 0.80 and average compression rate of 0.88% for video summary.