Very loopy belief propagation for unwrapping phase images

摘要: Since the discovery that best error-correcting decoding algorithm can be viewed as belief propagation in a cycle-bound graph, researchers have been trying to determine under what circumstances "loopy propagation" is effective for probabilistic inference. Despite several theoretical advances our understanding of loopy propagation, knowledge, only problem has solved using on Gaussian channels. We propose new representation two-dimensional phase unwrapping problem, and we show produces results are superior existing techniques. This an important result, since many imaging techniques, including magnetic resonance interfer-ometric synthetic aperture radar, produce phase-wrapped images. Interestingly, graph use very large number short cycles, supporting evidence minimum cycle length not needed excellent propagation.