In this paper, based on the principle that signals only occurs on a few time-bins (timing accuracy of a modern TCSPC module) due to short period illumination and the noise which is almost a constant for all bins, we designed an algorithm for denoising by using signals from multiple detectors and further improving the performance by adopting signal processing technique (e.g. median filtering). To exploit the information from the histograms efficiently, it is necessary to choose the gating width in an appropriate range, so that a well-placed window is large enough to capture most of signals, but without accepting too many noise. Whether to accept the signal is determined by weighting signals from multi-detectors within each gate width, where only multiple detections are recorded during the period can be accepted, otherwise detection is considered as noise and removed in the final reconstruction. Thus the three-dimensional information (the twodimensional spatial information and the depth information) of the target can be obtained accurately. In the algorithm, the results are evaluated numerically by signal to reconstruction error (SRE). The advantage is learned by comparing with several classic methods, the data used in this paper are derived from a mathematical model which account for Poisson statistics and Lidar equation. © 2019 copyright SPIE. Downloading of the abstract is permitted for personal use only.