Unwrapping phase images by propagating probabilities across graphs

摘要: Phase images are derived from source by applying a modulus operation to each pixel value. unwrapping is the problem of inferring original, unwrapped values wrapped values, using prior knowledge about smoothness image. One approach solving this infer gradient vector field image and then integrate field. The in particular direction at equal observed difference plus an unknown integer number shifts. We introduce technique for these shifts low-complexity probability propagation algorithm, applied graphical model that prefers match phase constrains satisfy properties present results region Sandia National Laboratories.