Spectral edge detection in two dimensions using wavefronts (Q617705)

This paper presents a multidimensional edge detection algorithm using the wavefront (i.e., a set of surfels of the form \((x,\vec{k})\) where \(x\) and \(\vec{k}\) represent a position and a direction, respectively) in order to precisely identify the singularities (or discontinuous edges) in images obtained in the spectral (or Fourier) domain. The wavefront extraction using directional filters with suboptimal parameters is summarized as the surfel extraction algorithm after theoretically intense and rigorous derivations and proofs. The performance of the proposed segmentation algorithm, which is based on connecting the extracted surfels, is tested for the low-resolution (and possibly noisy) MRI image, and its result illustrates excellent recovery of discontinuous edges in the image. However, since the theoretical foundation of this work is dependent on some assumptions made by the authors, there can be limitations. The authors relegate the extensions of this method for the future work.