Airborne Lidar can provide three dimensional point clouds in a very efficient manner. Most operational systems receive one or more discrete echoes from each emitted laser pulse. Full-waveform lidar systems record much dense return signals of each pulse. This facilitates further analysis with the waveform in the post processing. With this option, additional information of the target's properties could be extracted, as well as improving the quality of the extracted features. Both digital terrain model (DTM) and canopy height model (CHM) generation can be benefited from this feature. This study investigates different echo extraction schemes, namely, center of gravity estimation (COG), Gaussian pulse fitting (GPF), Gaussian pulse estimation (GPE), and two classification methods, simple decision tree and simple classification procedure, with a real dataset collected with RIEGL LMS-Q560. The differences between these schemes are comparatively studied.