Video++: An object-oriented approach to video algebra.

Abstract

Video has become an important medium in the field of multimedia communications. An increasing amount of video processing software is being developed to manipulate video (and audio) in an efficient and easy-to-use manner. Great advances have been made in the fields of video compression, video indexing, and content-based retrieval but, in spite of its growing importance, video has still not been endorsed as a data type in the software development community. The main reason is that video, unlike other data types such as integers and strings, is complex and difficult to manipulate. Its size prohibits it from being loaded into the random access memory of most computers, its spatial structure is complex, and the temporal order of its frames must be preserved at all times. This thesis proposes a new concept, video algebra, which allows developers to manipulate video as basic data types. This thesis also presents a new method for structuring video which is used by video algebra. A prototype has been developed in the form of an object-oriented framework called Video++. Video++ introduces two new concepts; the video markup language (VML) and the video description header. (Abstract shortened by UMI.)