To support video applications efficiently, video objects must be modeled and structured in secondary storage effectively, allowing flexible, easy, and efficient retrieval as well as support for dynamic video application development. In this paper, we describe a video management framework called VideoMAP (which stands for Video Management and Application Processing). The VideoMAP framework provides three components: the Segmentation and Clustering Component (SCC), the Semantic Modeling Component (SMC), and the Query Processing Component (QPC). SCC structures video objects through video partitioning, keyframe selection and segment clustering. SMC organizes video segments/clusters into semantically meaningful hierarchies with annotated feature indices. QPC provides flexible query specification and processing, and supports for external video application development. The framework employs a layered architecture, providing a flexibility and extensibility for each component to incorporate different schemes for different kinds of video objects. A prototype of VideoMAP is currently being constructed using an object-oriented database toolkit on the PC platform.