IMAGE DEMODULATION USING MULTIDIMENSIONAL ENERGY SEPARATION

Locally narrow-band images can be modeled as two-dimensional (2D) spat ial AM-FM signals with several applications in image texture analysis and computer vision. We formulate an image-demodulation problem and pr esent a solution based on the multidimensional energy operator Phi(f) = parallel to del f parallel to(2)-f del(2)f This nonlinear operator i s a multidimensional extension of the one-dimensional (1D) energy-trac king operator psi(f) = (f')(2)-ff '', which has been found useful for demodulating 1D AM-FM and speech signals. We discuss some interesting properties of the multidimensional operator and develop a multidimensi onal energy-separation algorithm to estimate the amplitude envelope an d instantaneous frequencies of 2D spatially varying AM-FM signals. Exp eriments are also presented on applying this 2D energy-demodulation al gorithm to estimate the instantaneous amplitude contrast and spatial f requencies of image textures bandpass filtered by means of Gabor filte rs. The attractive features of the multidimensional energy operator an d the 2D energy-separation algorithm are their simplicity, efficiency, and ability to track instantaneously varying spatial-modulation patte rns.