Fast Algorithm for Video Shot Boundary Detection Using SURF features
Eman Hato, Matheel Emaduldeen Abdulmunem · 2019
This paper presents an efficient and fast algorithm to detect abrupt shot transition automatically. The proposed algorithm consists of three - steps: feature extraction, features matching, and similarity calculation. The Speeded-Up-Robust-Features (SURF) features are extracted from half number of frames of video file to reduce the execution time. Next, features matching is performed between two features vectors to determine which is the nearest neighbors between features vectors using the distance function. The similarity is computed using a number of matching features and then compared to a predefined global threshold for the abrupt shot detection. The results demonstrate that the proposed algorithm achieves 0.9984 recall, 0.977 precision and 0.9991 F-measure. The proposed algorithm reduces the execution time while still detecting abrupt transitions at high performance rate.