Robust video retrieval using temporal MVMB moments

Abstract

In this paper, we propose motion pattern-based descriptor which exploits both spatial and temporal features to characterize video sequences in a semantics-based manner. The Discrete Cosine Transform (DCT) is applied to convert the high-level features from the time domain to the frequency domain. The energy concentration property of DCT allows us to use only a few DCT coefficients to precisely capture the variations of moving blobs. In comparison with the frequently used motion activity descriptors, the RLD and SAH of MPEG-7, the proposed descriptor yields 38% and 19% average gains over RLD and SAH, respectively.