Abstract: When measuring step samples by using microscope, a wealth of step information of threedimensional shape itself is highly vulnerable to be interfered by highfrequency noise. How to maintain the step characteristics of threedimensional shape while filtering noise and achieve highprecision measurement of threedimensional shape of the sample surface information is an important research question. We did the research on wavelet denoising method based on modulus square threshold method for step threedimensional shape information by using good spacedomain and frequencydomain localization properties of wavelet function. Haar wavelet was selected for the step characteristics of the sample. The method was applied in the threedimensional height profile which was obtained through laser differential confocal microscope developed by our research group. The height measurement result of the sample after denoising is consistent with the scanning result of OLYMPUS confocal microscope. The deviation is 0.146 8 nm. It can satisfy the requirement of the followup measurement analysis of threedimensional shape information and prove the effectiveness of the algorithm.