Analysis of Iterative Signal Deconvolution Algorithms and Their Sensitivity to Noise Using Simulated Full-Waveform LiDAR Data

Full-Waveform lidar provides detailed insights into vegetation and terrain structures by capturing the complete return signal, but deconvolution is often required to correct for system contributions and recover the true backscattering response. This study compares two frequently used deconvolution algorithms namely Richardson-Lucy (RL) and Gold deconvolution algorithm, along with their boosted variations, for their noise robustness, peak recovery accuracy, and shape preservation. Using physics-based simulated full-waveform lidar data generated through the “Digital Imaging and Remote Sensing Image Generation” radiative transfer model, the algorithms were tested under three scenarios: (i) noiseless waveforms and system contributions, (ii) noisy waveforms with noiseless system contributions, and (iii) noise affecting both. Results show that RL is comparable to Gold in noiseless conditions; however, Gold consistently outperforms RL in noisy scenarios due to its greater noise resilience, faster computation, and superior shape recovery. These findings highlight Gold deconvolution algorithm as more robust choice for noise-affected remote sensing applications.