Normalization of fringe patterns using the bidimensional empirical mode decomposition and the Hilbert transform.

We evaluate a data-driven technique to perform bias suppression and modulation normalization of fringe patterns. The proposed technique uses a bidimensional empirical mode decomposition method to decompose a fringe pattern in a set of intrinsic frequency modes and the partial Hilbert transform to characterize the local amplitude of the modes in order to perform the normalization. The performance of the technique is tested using computer simulated fringe patterns of different fringe densities and illumination defects with high local variations of the modulation, and its advantages and limitations are discussed. Finally, the performance of the normalization approach in processing real data is also illustrated.