Abstract : Motile cilia play a major role in fluid circulation in many biological systems such as the oviduct, the airway and the cerebral ventricles and defects in cilia beating lead to pathologies called ciliopathies. The estimation of the cilia beating frequency (CBF) is of great interest to understand how the CBF modulates the liquid fluxes and how CBF is controled by the intra- and/or extra-cellular medium composition of the ciliated cell in a physiological process. Such estimation is almost impossible to be computed on a single cilium because of overlapping patterns. In addition, motion artifacts and camera defaults may hinder or bias the computation of the frequency variations during long lasting experiments. In this paper is provided a methodological development matching the cilia properties. It is first based on a grid removal and a shift compensation. Second, an harmonic model of the observed cilia through image recordings is proposed within a Maximum Likelihood Estimator framework. It is proven that in contrast to classical approaches the best estimation of the frequency is obtained by averaging the squared Fourier trans- form of individual pixels followed by a particular summation over the frequency, namely the Compressed Spectrum. Illustrative examples show the performances of this robust method compared to classical ones.