A Local Phase Modulated Active Contour for Aerial and Natural Image Segmentation

Varying illumination and poor contrast causes inhomogeneity in real world images. It is difficult to segment such images having intensity inhomogeneities. Therefore proposed method utilized the moments of phase congruency (PC) in the energy functional of active contour model to segment such images. Both maximum and minimum moments of phase congruency help to detect edges and corners simultaneously, rather than only edges or corners. Thus active contour can easily stop the evolution at the desired edges. Satellite images generally are both noisy and inhomogeneous. Therefore promising results in this paper demonstrate the potential of proposed approach for extracting buildings and/or map from the aerial and satellite images. Experimental results on synthetic and natural images further show the robustness toward inhomogeneity and noise of proposed model.