By analyzing the on-board lightning data of the FY-4 lightning mapping imager (LMI), there are three types of nonlightning events: ghost noise, high-energy-particle-trajectory noise, and shot noise. To improve the performance of lightning detection, based on an in-depth analysis of the characteristics of false events, the ghost-image-filtering method based on spatial energy correlation, the high-energy-particle-trajectory noise-filtering method based on Hough linear detection, and the shot-noise-filtering method based on spatiotemporal correlation clustering are proposed, and the lightning false-alarm-filtering-processing system method is performed. To test the performance of the false-alarm-filtering algorithm, the lightning-filter algorithm is applied to process two thunderstorms in Yunnan and Hainan provinces. We compare the lightning data obtained by the LMI with those obtained by ground-based network. The results show that the distribution characteristics of lightning events processed by the proposed algorithm are consistent with the ground-based detection results, which verifies the continuous tracking performance of the FY-4 lightning mapping imager for the lightning change process.