A Novel Framework for Semantic-Based Video Retrieval

Xiaoming Nan,Zhicheng Zhao,Anni Cai,Xiaohui Xie
DOI: https://doi.org/10.1109/icicisys.2009.5357647
2009-01-01
Abstract:In this paper, a novel framework for semantic-based video retrieval is proposed. 15 low-level visual features on different levels are extracted and a supervised SVM classifier is trained for each feature. We have explored early fusion schemes between SIFT and SURF, and evaluated 4 kinds of later fusion strategies. Experiments on TRECVID dataset show that the proposed system is effective and stable.
What problem does this paper attempt to address?