Not yet reviewed by Pith; the record is open.

We develop and validate a new algorithm called \textit{primary track recovery} (\ptr) that effectively deconvolves known physics and detector effects from nuclear recoil tracks in gas time projection chambers (TPCs) with high-resolution readout. This gives access to the primary track charge, length, and vector direction (helping to resolve the "head-tail" ambiguity). Additionally, \ptr provides a measurement of the transverse and longitudinal diffusion widths, which can be used to determine the absolute position of tracks in the drift direction for detector fiducialization. Using simulated helium recoils in an atmospheric pressure TPC with a 70:30 mixture of He:CO$_2$ we compare the performance of \ptr to traditional methods for all key track variables. We find that the algorithm reduces reconstruction errors, including those caused by charge integration, for tracks with mean length-to-width ratios 1.4 and above, corresponding to recoil energies of $20$ keV and above in the studied TPCs. We show that \ptr improves on existing methods for head-tail disambiguation, particularly for highly inclined tracks, and improves the determination of the absolute position of recoils on the drift axis via transverse diffusion. We find that \ptr can partially recover charge structure integrated out by the detector in the $z$ direction, but that its determination of energy and length have worse resolution compared to existing methods. We use experimental data to qualitatively verify these findings and discuss implications for future directional detectors at the low-energy frontier.