Abstract
This paper presents a novel framework to detect shot boundaries based on the One-Class Support Vector Machine (OCSVM). Instead of comparing the difference between pair-wise consecutive frames at a specific time, we measure the divergence between two OCSVM classifiers, which are learnt from two contextual sets, i.e., immediate past set and immediate future set. To speed up the processing procedure, the two OCSVM classifiers are updated in an online fashion by our proposed multi-instance incremental and decremental one-class support vector machine algorithm. Our approach, which inherits the advantages of OCSVM, is robust to noises such as abrupt illumination changes and large object or camera movements, and capable of detecting gradual transitions as well. Experimental results on some benchmark datasets compare favorably with the state-of-the-art methods.