Integrated Video Object Segmentation and Shape Coding

To enable content-based access and manipulation of video content, modern multimedia communications require per-object based access to video data. Two additional building blocks of video transmission systems needs to be defined, which enable access to video objects at the receiver end. These are video segmentation, which decomposes video frame into a set of layers, where each layer includes segmentation mask of one semantic video object, and shape coding system, which enables transmission of the segmentation information to the receiver. This paper presents a novel, integrated approach to the fore mentioned problem, where the parameters of object segmentation module and shape coding module are dynamically adjusted depending on bandwidth constraints. Background differencing with morphological filtering is used for extracting shape information, and B-splines are used for shape approximation. An efficient predictive scheme for adaptive arithmetic encoding of spline control points is used for shape coding. Key-Words: video coding, shape coding, video object segmentation