A bispectral-based test for spatial reversibility with application to texture images

Statistical approaches to image modeling have largely relied upon random models which characterize the 2-D process in terms of its first- and second-order statistics, and therefore cannot completely capture phase properties of random fields which are non-Gaussian. This constrains the parameters of noncausal image models to be symmetric and therefore the underlying random field to be spatially reversible. Higher- than second-order statistics are used to derive and implement a spatial reversibility test to validate this modeling assumption and consequently determine if higher-order statistics would be more appropriate for modeling an image. The performance of this test is analyzed both theoretically and experimentally and then applied to texture images