Variable Precision Fuzzy Hit-or-Miss Transformation Models to Object Identification in Grey-Scale Images

摘要

Morphological image operators are important geometry-based image transformations, whereas fuzzy set theory and fuzzy logic, favoured for their ability to describe and cope with imprecise and uncertain natures of objects, are successfully applied to image processing and pattern recognition. In this paper, generalizations of hit-or-miss transformation (HMT) for grey-scale images are investigated in the framework of fuzzy set theory and fuzzy logic. To address the problem of robust object identification, a notion of fuzzy vertical translation is proposed and two variable precision fuzzy HMT models are developed for grey-scale images to cater for flexible selection of structuring elements in the fuzzy setting. Their extensions to the rank case are designed to adapt the HMT to noisy images and images with inhomogeneous intensities. Superior performance of the proposed models over the state-of-the-art HMT models in objection extraction is verified by a series of experiments on synthetic images and real images.