This paper proposes a three-dimensional (3-D) phase correlation-based rigid volume registration method for estimation of the lung tumor motion in four-dimensional (4-D) computed tomography (CT). In the lung cancer radiotherapy treatment, the 4-D thorax CT has been used to observe the spatiotemporal tumor motion with the patient's respiration. In the proposed method, the respiration-induced lung tumor motion is approximately modeled as a 3-D translation between two CT volumes obtained from two respiratory stages. A 3-D phase correlation of the two CT volumes is calculated by using the 3-D inverse discrete Fourier transform (IDFT) of the normalized cross power spectrum of the two volumes. The tumor motion values along three directions are estimated by locating the highest peak of the 3-D phase correlation. The experimental results show that the proposed method is capable of estimating the tumor motion with high accuracy and low computational cost.