Morphological image segmentation by local monotonicity

Image segmentation is fundamental task in image processing and multimedia. We propose a general method for image segmentation based upon the relationship between mathematical morphology and the local monotonicity of signals. In this paper, we introduce a two-dimensional generalization of the concept of local monotonicity based on mathematical morphology. A multiscale representation of an image is morphologically generated, wherein the degree of local monotonicity determines scale. Then, a morphological edge detection operator exploits the monotonicity property in performing segmentation. The segmentation allows specification of object scale, edge detail, and contrast and is applicable to object-based video coding.