Phase Congruency

Abstract

This paper presents a frequency-domain analysis tool called phase congruency. An overview of what is required from phase congruency, its purpose in fusion and the goals provides a detailed formulation of phase congruency; each step is discussed in the context of multi-spectral information and the goal of extracting a stable measure to compare the disparate modalities. An investigation which aims to identify predictable quantities extracted in the phase congruency analysis process is specified and carried out. it clarifies the objectives of this investigation and the approach to achieving them. . Thermal images appear very different from the familiar visible spectrum, material boundaries and object edges still make elements in the scene recognisable to the human eye. This observation motivates the use of structural features (e.g. edges, contours and corners) as stable and matchable elements of multi-spectral scenes. Edges are commonly extracted using the Canny edge detector in which linear filtering with a heuristic threshold is used to detect edges at points of high contrast, i.e. a large gradient magnitude between adjacent pixels is used to indicate the presence of an edge. The undesired effect of this method is that areas of low contrast are left devoid of features and, due to the xed-size filter, blurred edges are poorly localized  Thermal images in particular often contain large areas of low contrast, and edges are often blurred due to the low spatial resolution and thermal phenomena .Therefore, structural features cannot be repeatable detected or compared using conventional gradient-based methods.