Three novel edge detection methods for incomplete and noisy spectral data (Q734948)

In the present paper, conjugate concentration kernels and sparse representations are utilized to recover edges from spectral information of piecewise smooth data. The authors propose the randomly\,-\,based sparse inverse fast Fourier transform, the randomly\,-\,based edge detection method, the total variation\,-\,based compressed sensing method, and an improved zero-crossing method as novel methods to detect edges from both complete and incomplete spectral data, which are possibly contaminated with noise. Some numerical results are provided to illustrate the relative advantages of these novel methods as effective edge detectors in practical applications.