Robust filters for low-level vision

摘要

The first stage in many early vision problems consists in filtering the input image. In this paper we show how to extend the classical results on robust smoothing based on Bayesian estimation with prior Markov random field models to the case of complex-valued pass-band (quadrature) filters. Examples of the application of these results to stereo reconstruction, texture segmentation and fringe pattern analysis in optics are also presented.