Object-scalable dynamic coding of visual information

Abstract

This paper describes an extension of a dynamic video coding scheme to provide object scalable functionalities. As a particular instance of the dynamic coding concept, the coding scheme considered here jointly optimizes video data partition and representation modes. Indeed, as there exists no universal video coding method, dynamic coding insures the choice of the most efficient technique (for instance DCT, fractal or motion compensation) for each data segment. These data segments are themselves optimally partitioned within the original frame (respectively region of interest). An optimization algorithm achieves this joint data partition/representation modes selection to yield the best rate/distortion compromise within the available set of possible solutions, under a rate or distortion constraint. Such a dynamic coding algorithm designed for low bitrates was proposed to MPEG-4 first set of tests in November 1995. This paper describes the corresponding object scalable coding scheme.