Method for automatic extraction of expressive elements from motion pictures for video synthesis, training, archival, and search
Original Publication Date: 2002-Dec-15
Included in the Prior Art Database: 2003-Jun-21
Disclosed is a unique computational method for extraction of expressive elements in motion pictures for deriving high level semantics of stories portrayed, enabling better video annotation and interpretation systems. Though a great deal of work has been done in low level content based analysis of video, the area of automatic semantic analysis or high level interpretation of video is just beginning. Our approach described here differs from many recent approaches in that while previous approaches have sought to model very specific events in a specific domain, our scheme attempts to understand the `expressiveness'' of the medium and the thematic units (high-paced section, tranquil rhythm etc.) naturally underlined by different expressions that are pervasive regardless of the domain of the content. Our method is motivated and directed by the existing cinematic conventions known as film grammar, and attempts to extract concepts such as tempo, rhythm, tone, mood, and genre of a film. As a first step of this broad approach, it uses the attributes of motion and shot length to define and compute a novel measure of tempo or pace of a movie. Our method shows tempo as a useful attribute in its own right in automatic video annotation systems, and a promising component of higher semantic units such as tone or mood of a film.