In the millimeter wave imaging of personal security inspection， it is with low probability to detect the dangerous and violent/explosive things in the limited two-dimensional imaging plane. Therefore， the human result reconstructed by three-dimensional fully focused algorithm is demanded urgently. It is difficult to achieve a balance between the accuracy and efficiency for the conventional three-dimensional reconstruction algorithms that are purely either in time or frequency domain. To this end， time-frequency coordinated three-dimensional reconstruction algorithm （TFC-3DR） is proposed based on the dimensional decomposition and algorithmic coordination scheme， in which the efficiency of frequency-domain algorithm and the accuracy of time-domain algorithm can be combined and coordinated. In order to guarantee the operational efficiency， the range migration is uniformly calibrated in the range-height plane under the azimuth invariance characteristics of the range migration algorithm （RDA） in the frequency-domain. Then， the time-domain back projection （BP） is employed in the range-angle plane through point-by-point coherent accumulation of pulse-compressed data and resolution units to guarantee the imaging accuracy. In the end， TFC-3DR is established with both high efficiency and accuracy. Due to the introduction to BP algorithm， the scanning trajectory of the array antenna in the angle dimension is flexible and variable， which reduces the blind area of the human body， so that highly missed and false detection rate can be avoided. Simulated and practical data of human body is performed to verify the effectiveness of the proposed algorithm. Meanwhile， the superiority over the proposed algorithm is examined by comparing with the BP reconstruction algorithm in terms of the reconstruction accuracy and computation complexity.