In an attempt to find parameters of a time series which are absolutely robust with respect to nonlinear distortion, we introduce a function called the entropy profile which measures in some sense the distance between the given process and white noise. This concept combines a clear definition and a simple algorithm, which apply to arbitrary stationary time series, with an informative graphical representation similar to the Fourier spectrum. For sequences derived from one-dimensional maps, the entropy profile indicates periodic and almost periodic behavior and the presence of Markov partitions.