Parameter estimation of 2D polynomial phase signals using NU sampling and 2D CPF

The two- dimensional ( 2D) cubic phase function ( CPF) is known as a highly accurate 2D polynomial phase signal estimator, but it has limited applicability due to the requirement for the 3D search for second- order partial phase derivatives. The authors propose an interpolation- based approach simulating non- uniform ( NU) signal sampling in order to reduce the 2D CPF calculation complexity. The NU resampling enables the 2D CPF evaluation using the 2D fast Fourier transform and searches over mixed- phase parameter. The computational complexity is reduced from O( N5) to O( N3 log2 N). The additional stage with dechirping, filtering and phase unwrapping is introduced to refine parameter estimates.