Shot Boundary Classification and Refinement Using Inter-Frame Similarity Patterns

In content-based video analysis, shot boundary detection (SBD) is a common first step to segment a video into fundamental units of shots, each composing consecutive frames filmed with a single camera act. Many methods have been proposed in the literature for SBD, and many require the use of different models to determine the types of various shot boundaries. In this paper, we propose a new and effective shot boundary classification (SBC) and refinement method by analyzing the inter-frame similarity patterns around the detected shot boundaries. In particular, the proposed method can classify the detected shot boundaries into five types-shot cuts, wipes, dissolves, fades, and non-boundaries-using a more generic and accurate manner, and the non-boundaries obtained in the classification can be used to refine the SBD results by eliminating the false detections to attain a better detection performance, as evidenced by the experimental results presented