CONTENT-BASED VIDEO SHOT BOUNDARY DETECTION USING CANNY EDGE FEATURES

Abstract

Video Shot Boundary Detection (VSBD) is a vital preliminary step in content-based video analysis, enabling efficient indexing, retrieval, summarization, and editing workflows. This paper introduces a robust and computationally efficient method for detecting shot boundaries especially abrupt cuts using the Canny edge features. Our approach employs the following pipeline: frames are converted to grayscale and Canny edge features are extracted. Normalized ECR is computed on consecutive frames, then threshold is set and shot transitions are flagged. We benchmark our method on standard datasets such as TRECVID, evaluating performance in terms of precision, recall, F1-score. Our proposed method average F1-score 93.17.