KFE-OPPSO: A New Key Frame Extraction Technique with Orthogonal Polynomials and Particle Swarm Optimization

Abstract

In this paper, a new Key frame Extraction technique with Orthogonal Polynomials and Particle Swarm Optimization (KFE-OPPSO) is proposed. The input video file is converted in to frames and is applied with orthogonal polynomials to derive transform coefficients. These transformed coefficients are then utilized to detect edges in a frame by identifying zero crossings in second directional derivatives. These edge detected frames are then applied morphological operation “dilate” followed with “invert” operation to yield clear and visible edges. Further, an edge movement in video frames is performed between current inverted frame and previous edge detected frame. Finally, these frames are compared with each other to detect different types of shots such as hard cuts, fade and dissolve effects in terms of Edge Change Ratio (ECR).  From the detected shots, color feature is extracted with few orthogonal polynomials coefficients and then fed to Particle Swarm Optimization (PSO) for selecting optimal key frames. Experiments are carried out on 20 different videos downloaded from the standard open video project. The experimental results indicate that the proposed scheme achieves better precision, recall and f-measure rate than other existing systems in terms of key frames selection.