Modeling of ultrasound speckle with application in flaw detection in metals

RF ultrasound backscatter speckle is modeled by a parametric noncausal Gaussian Markov field that appears to capture the speckle structure well, is consistent with the first-order marginal statistics of the complex amplitude speckle, and explains the homogeneous coarseness appearance of the speckle in terms of a homogeneous colored random field with a definite correlation structure. The Markov property results in the compact parametrization of the backscatter echo field by a few parameters, and constitutes a huge reduction in the representation and analysis complexity of the backscatter image. Use is made of the intensity as well as phase information, associated with the acoustic image by analyzing and modeling the acoustic RF data rather than the intensity data. The additional phase information associated with the RF data allows for the statistics of the complex backscatter echo field or, equivalently, that of the RF field to be completely specified. The approach provides a coherent theoretical basis for the design of statistical tests for detecting targets embedded in the speckle image, for the case where the speckle model parameters are either a priori known or unknown.<
>