Segmentation-based image coding by morphological local monotonicity

Abstract

Segmentation-based image and video coding is desirable for many multimedia applications due to the additional functionality provided by object-based representation. Methods of object-based coding have generally treated the segmentation and encoding processes as separate problems. Here, we present an integrated segmentation and coding method unified by the theoretical structure of morphological local monotonicity. This unified segmentation/coding scheme utilizes morphological operators within a nonlinear scale-space to generate a segmentation. The segmented regions are independently coded and reconstructed using morphological generalization of Laplace's equation in a multiresolution framework. The coding procedure is appropriate for non-textured imagery and avoids arbitrarily chosen constants. Examples are given for two-dimensional grayscale imagery.