We propose modified local mean decomposition to analyze signals of fringe pattern adaptively. It decomposes the signals into a set of functions, each of which is the product of amplitude signal and frequency signal. Then, the physical components of noise and background are extracted according to the corresponding product functions. Moreover, to solve the most likely mode mixing problem, a high-frequency signal is constructed according to the amplitude and frequency characteristics of the intrinsic noise. Using the presented method, carrier signals are recovered accurately as well as the wrapped phase. Compared experiments illustrate the validity of this method.