The axial resolution of confocal microscopy is not only dependent on optical characteristics but also on the utilized peak extraction algorithms. In previous evaluations of peak extraction algorithms, sample surface height is generally assumed to be zero, and only sampling-noise-induced peak extraction uncertainty was analyzed. Here we propose a sample surface-height-dependent (SHD) evaluation model that takes the combined considerations of sample surface height and noise for comparisons of algorithms’ performances. Monte Carlo simulations were first conducted on the centroid algorithm and several nonlinear fitting algorithms such as the parabola fitting algorithm, Gaussian fitting algorithm, and ${\mathrm{sinc}}^2$ fitting algorithm. Subsequently, the evaluation indicators, including mean peak extraction error and mean uncertainty were suggested for the algorithms’ performance ranking. Finally, experimental verifications of the SHD model were carried out using a fiber-based chromatic confocal system. From our simulations and experiments, we demonstrate that sample surface height is a critical influencing factor in peak extraction computation in terms of both the accuracy and standard deviations. Compared to the conventional standard uncertainty evaluation model, our SHD model can provide a more comprehensive characterization of peak extraction algorithms’ performance and offer a more flexible and consistent reference for algorithm selection.