Small-animal positron emission tomography (PET) systems are widely used for pathological and pharmacological studies. For good image quality, the timing resolution should be kept as low as possible, so we proposed a pulse reconstruction method based on curve fitting using a bi-exponential model for digital timing pickoff. To evaluate the proposed method, five different digital timing algorithms, leading edge discrimination (LED), digital constant fraction discrimination (DCFD), maximum rising interpolation (MRI), initial rising interpolation (IRI), and pulse reconstruction (PR) were established and compared. A single gamma pulse obtained from the silicon photomultiplier (SiPM) was simulated and the timing resolution was calculated according to the algorithms. Additionally, these methods were applied to the real PET dataset obtained from the AGPET detector and the timing resolution was derived. The simulation results proved that the PR method showed constant timing resolution independent from pulse rise time and sampling rate. In the experimental results, the PR method showed the best timing resolution of 0.9 ns to compare other digital methods. The results demonstrated that the proposed PR method can improve the performance of the small-animal PET with a relatively slow ADC.