Video Copy Detection Based on Principal Component Analysis

Abstract

Content-based video hashing was proposed for the purpose of video copy detection. Conventional video copy detection algorithms apply image hashing algorithm to either every frame or key frame which is sensitive to video variation. In our proposed algorithm, key frames including temporal and spatial information are used to video copy detection, Discrete cosine transform (DCT) is done for video key frame and feature vector is extracted by principal component analysis ( PCA ). An average true positive rate of 99.31% and false positive rate of 0.37% demonstrate the robustness and uniqueness of the proposed algorithm. Experiments indicate that it is easy to implement and more efficient than other video copy detection algorithms.