Abstract The deconvolution method with the TV (Total Variation)-Tikhonov algorithm and the MRI (atomic Mixing-Roughness-Information depth) depth resolution function was recently proposed for reconstructing the original in-depth distribution of composition directly from measured depth profiling data. In this paper, it shows that the two parameters of α and β in the TV-Tikhonov algorithm play an important role in determining the deconvolution result with high precision and strong anti-noise performance. Because the measured depth profiling data often have some kinds of noise and/or insufficient data points, the optimization of α and β values is necessary and is carried on by the genetic algorithm. The convergence of the genetic simulation and the influences of these two parameters on the deconvolution result are evaluated quantitatively. Finally, as an example, the original layer structure of a Ni/Ag multilayer is deconvoluted directly from the measured AES depth profiling data and is agreed well with the TEM measurement.