Purpose <p>Radiation detector signals are processed by preamplifiers to produce amplified and stretched pulses. However, this also increases the likelihood of pulse overlap, known as pile-up. Digital processing modules have enabled the unfolding-synthesis strategy for pulse processing, where digitized preamplifier output pulses are deconvoluted into impulse signals, then formed to the shape of interest, in order to reduce the pile-up effect.</p> Methods <p>The present study compares several unfolding algorithms for mono- and bi-exponential pulses. For smoothing signals, two strategies are considered: a successive application of moving mean and moving median filters, and Savitzky–Golay filtering. Both simulated and experimental pulses from an NaI(Tl) detector are taken into account. Savitzky–Golay filtering is shown to enhance pulse amplitude (energy) resolution but offers weaker noise filtering than the mean–median strategy.</p> Results and Conclusion <p>Results show that algorithms based on the bi-exponential pulse assumption produce signals closer to impulse, in order to preserve more pulses at the expense of pulse amplitude resolution, while those based on the mono-exponential assumption generally give signals of width at least equal to the rise time of the input pulse. A proposed bi-exponential-based algorithm introduces a tunable term for a balanced solution.</p>