Improving Phase-Congruency Based Feature Detection through Automatic Scale-Selection

摘要

In this paper we present a novel method for computing phase-congruency by automatically selecting the range of scales over which a locally one-dimensional feature exists. Our method is based on the use of local energy computed in a multi-resolution steerable filter framework. We observe the behaviour of phase over scale to determine both the type of the underlying features and the optimal range of scales over which they exist. This additional information can be used to provide a more complete description of image-features which can be utilized in a variety of applications that require high-quality low-level descriptors. We apply our algorithm to both synthetic and real images.