The Maximum Entropy Method is widely used for reconstruction of real non-negative functions, such as images (intensity distributions) in optics and astronomy. The problem of reconstruction exists not only for real non-negative functions: in radio holography, for example, it is often necessary to reconstruct a coherent source field distribution described by a complex function. In this paper the Generalization of Maximum Entropy Method for reconstruction of functions of different types (real non-negative as well as real with alternating signs and complex ones) is suggested. Though this problem is considered for two-dimensional functions it is evident that the generalization obtained can be applied for functions of different dimensions. Numerical simulation results show high quality of reconstruction of complex functions and stability of the algorithm in the presence of measurement errors.