Curvelet decomposition for detection of cylindrical targets

Curvelets are a multiscale system with very high directional sensitivity. A new detection algorithm is herein described which operates on a curvelet decomposition of acoustic imagery. The algorithm detects the presence of cylindrical targets through a statistical mapping of curvelet coefficients. The coefficients are calculated as an inner product between image features and a curvelet basis element. The similarity in appearance between cylindrical targets and curvelet basis elements yield an accurate detection algorithm with a very low false alarm rate.