The performance of the three-dimensional reconstruction algorithm in Geiger-mode Avalanche Photo Diode (Gm-APD) laser imaging LiDAR, particularly in terms of target fidelity and image signal-to-noise ratio (SNR), is highly dependent on the number of statistical frames used. Generally, increasing the number of frames enhances reconstruction accuracy but compromises real-time imaging capability. Conversely, using fewer frames can lead to target omission and reduced image SNR. To address the challenge of accurate 3D reconstruction under low-frame conditions, a novel algorithm based on a histogram enhancement strategy is proposed. The method begins by employing the Pearson correlation coefficient to sum histograms within correlated neighborhoods, thereby enhancing signal features. A sliding window search is then applied to estimate the target distance range, effectively filtering out noise and yielding an enhanced histogram. Subsequently, a low-complexity bi-parameter maximum likelihood estimation is used to reconstruct the distance image. Finally, spatial domain thresholding is applied to suppress residual noise and further improve image quality. Experimental results demonstrate that with only 50 statistical frames, the proposed algorithm improves target fidelity by 0.13 and enhances the image SNR by 1.63 dB compared to the matched filtering method. These results confirm the algorithm’s effectiveness for real-time Gm-APD LiDAR applications where a limited number of frames is required.