Virtual-object video compression

A virtual-object compression method is presented in this paper. Object based compression methods are inspired by the publication of the MPEG-4 standard. However, object-based compression methods require several non-trivial processes, many of which continue to pose challenges to many researchers. These processes include but are not limited to: robust multiple object identification and tracking, and shape and texture coding of arbitrarily shaped three-dimensional objects. Therefore, a virtual-object compression method is developed which relaxes the constraints of the many difficult problems that object-based compression methods pose. In the virtual-object compression method, an image sequence is parsed into spatial locations which contain motion and those which do not. Those portions of video which do not contain motion are averaged into one background frame, and those which do contain motion are group into a virtual-object. Both the background and virtual-object are coded independently from one another. This compression method shows an improved performance over 3D wavelet video compression.