Partition Lattice Operators for Extraction of Semantic Video Objects

We conceive the problem of multiple semantic video object (SVO) extraction as an issue of designing operators on a complete lattice of partitions. In this paper, we propose a framework based on accurate spatial partition generation and application of optimal extraction operators on the generated partitions. Under the framework, we introduce a spatio-temporal regional maximum likelihood operator for extraction purposes. Some theoretical properties of the operators are established. Experimental results show that our scheme is capable of successfully handling multiple SVOs in a variety of scenarios.