Video Sequence Matching Using Singular Value Decomposition

This paper proposes a novel signature based on singular value decomposition (SVD) for video sequence matching. By considering the input image as a matrix, a partition procedure is first performed to separate the matrix into non-overlapping sub-images of a fixed size. The SVD process then individually decomposes each partitioned sub-image into an singular value and the corresponding singular vector factorization. As a result, several dominant singular values are obtained for each sub-image, allowing the dissimilarity to be determined between the reference video clip and the query one. Experimental results based on receiver operating characteristics (ROC) curves confirm the effective performance of the proposed video signature.