Extraction of key frames is an essential and significant stage in any video analysis application, aimed at describing the video content precisely by removing the redundancy. An optimized key frame extraction technique based on the Binary Butterfly Optimization Algorithm (BBOA) is introduced. Here, the extraction of Color Coherent Vector (CCV) features from each frame of the video and the application of the BBOA algorithm for minimising redundancy and maximising diversity between the frames is carried out. The process is applied iteratively until the precise number of frames is selected. Further, the system is extended by proposing a Content-Based Video Retrieval (CBVR) system using the selected key frames, extracting multiple features from Grey Level Run Length Matrix (GLRLM) texture features and Dual Tree- Complex Wavelet Transform (DT-CWT) shape descriptors along with CCV features. Due to the multiple features, the feature vector size is huge, so to reduce its dimension, a hybrid Binary Particle Swarm Optimization-Butterfly Optimization Algorithm (PSO-BOA) feature selection method is applied. The experiment was conducted on the UCF101 dataset, and our proposed system outperformed with a compression rate, precision, recall rate, F1 score, and FRR of 0.985, 0.913, 0.78, 0.836, and 0.965, respectively, demonstrating the effectiveness of using the hybrid Optimization algorithm in improving the efficiency of the CBVR system.

Content-Based Video Retrieval (CBVR) Binary Butterfly Optimization Algorithm (BBOA) Particle Swarm Optimization-Butterfly Optimization Algorithm (PSO-BOA) Colour Coherence Vector (CCV) Grey Level Run Length Matrix (GLRLM) Dual Tree- Complex Wavelet Transform (DT-CWT)