Median-style filters for noise reduction in composition analyses

Abstract Several commonly applied techniques for atom probe data noise reduction are compared to signal processing techniques specifically designed to maintain interfacial integrity. A median filter and a double-window modified trimmed mean filter, along with several other filters, are applied to both a simulated atom probe composition profile and to real atom probe data. The various methods are compared using average profile compositions and variance measurements. The double-window modified trimmed mean filter is found to smooth the data in the slowly varying regions of the profile without degrading sharp interfaces.