SEGMENTATION-BASED VIDEO CODING: TEMPORAL LINKING AND RATE CONTROL

ABSTRACT This paper analyzes the main elements that a segmentationbased video coding approach should be based on so that it can address coding efficiency and content-based functionalities. Such elements can be defined as temporal linking and rate control. The basic features of such elements are discussed and, in both cases, a specific implementation is proposed. 1 INTRODUCTION Coding algorithms relying on a content-based representation are becoming increasingly popular. The main difference between the techniques reported in the literature is the relative importance they assign to the spatial or the motion information. In a first set of examples [17, 5, 15], regions are characterized by their spatial homogeneity, whereas in a second set [8, 2], motion information is used as the main homogeneity criterion. Finally, examples of segmentation techniques involving both criteria can be found in [3, 1]. This last approach offers a large number of advantages: the spatial homogeneity allows a very accurate definition of the regions and the processing of generic sequences because, for example, it can deal with the appearance of new objects. The motion homogeneity criterion is important to limit the number of regions and therefore the coding cost associated to the partition. This limitation is generally obtained by merging spatial regions that can be compensated together. Neighbor regions with homogeneous motion are usually related to the concept of objects. Such mergings can therefore allow an object representation of the scene. This representation opens the door to content-based functionalities [7]