MVO-SBD: A Hybrid Optimized Approach for Detecting Shot Boundaries Invariant to Motion and Illumination Effects

This paper presents a novel Shot Boundary Detection (SBD) method that identifies transitions and their boundaries between consecutive video shots using the Multi-Verse Optimizer (MVO) to optimize the weights of a Feed-Forward Neural Network (FNN). MVO is chosen for its strong exploration-exploitation capabilities and high precision in finding optimal solutions. A threshold-based strategy is applied to effectively label transition and normal frames. The novel MVOFNN model is tested using benchmark datasets to confirm its superiority over FNN with a superior F1 score. Comparative comparison proves that the presented approach beats the latest approaches and is thus an efficient and powerful solution to detect shot boundaries for video processing.