Authors: Hiroshi Arisawa, Kiril Salev, Takashi Tomii
Tags: 1998, conceptual modeling
Video is considered the most effective medium for capturing events. There are two major approaches of representing video contents: the structured modeling approach and stratification. They require video to be divided into simple semantic units on top of which some structures are built to express high-level video semantics. Text is used to describe the necessary semantic units. A major problem of such an approach is that the descriptions tend to be incomplete and subjective. In addition, segmenting video sequences and describing their semantic contents is a tedious task. In the present paper we define some low-level semantic primitives that can be detected automatically and are “semantics-free” from the viewpoint of event semantics. Further, we offer a methodology for describing event semantics and deriving it from the low-level primitives. We describe a prototype system based on the methodology and offer schema for its implementation based on the Real World Database. Our approach reduces the problem of video segmenting to simply performing a database query.Read the full paper here: https://link.springer.com/chapter/10.1007/978-3-540-49121-7_49
