In recent years, medical, biological and scientific fields have been benefited from phase information, which can reveal the hidden features of various objects. By solving the transport of intensity equation (TIE), the phase distribution of an object can be obtained from a series of intensity measurements. Assuming the solving process is correct, the reconstruction accuracy is depending on the distances between planes, the number of planes, and the level of noise. Increasing the number of planes or utilizing multi-frame de-noise algorithm could improve the reconstruction accuracy certainly, but neither of them is a time-efficient strategy. In this work, an optimum defocus planes selection (OPS) method is proposed for reconstructing high quality phase information by solving the transport of intensity equation. It is shown that the difference image between two symmetrical separated, lager defocused planes contains a lot of lower frequency components of the phase distribution and the higher frequency components can be easily observed in the difference image between two nearly focused planes. Based on the phase transfer function (PTF), our method estimate a more accurate frequency spectrum of the object phase distribution, which is combined with different frequency components from the stack of through–focus intensity images. Both the simulation and experimental results demonstrate that this optimum defocus planes selection method can give a computationally efficient and noise-robust phase reconstruction with higher accuracy and fewer defocus planes.