Morphological pyramids for multiscale edge detection

Abstract

In this paper, we present an edge detector that is based on a morphological pyramid (MP) structure. The algorithm utilizes a multiresolution pyramidal structure created by successive morphological filtering and subsampling of the original image. The boundaries detected at coarse scale representations of the morphological pyramid (MP) are used to guide the detection of discontinuities at higher resolution levels. The segmentation and resulting edge detection yielded by the MP is particularly effective in the presence of impulse noise. We provide results that demonstrate superior solution quality over standard fixed resolution detectors and over previous multiresolution approaches. Because of the low computational cost of the MP edge detector, it is suitable for video tracking, image and video compression, and real-time object recognition.