In this paper, the instantaneous frequency estimation of nonstationary signals is considered. The instantaneous frequency is estimated from the timefrequency representation where certain percent of the coefficients is missing. The time-frequency representation is considered as an image, whose missing pixels are reconstructed by using compressive sensing recovery algorithms. As a time-frequency representation, the S-method is considered. The Compressive Sensing as a intensively growing novel approach for signal acquisition, ensures accurate signal reconstruction from relatively small percent of available information about the signal. The theory is verified by experimental results.