An unbiased ray-marching transmittance estimator

We present an in-depth analysis of the sources variance in state-of-the-art unbiased volumetric transmittance estimators, and propose several new methods for improving their efficiency. These combine to produce a single estimator that is universally optimal relative prior work, with up orders magnitude lower at same cost, has zero any ray non-varying extinction. first reduce truncated power-series estimators using novel efficient application U-statistics. then greatly average expansion order power series redistribute density evaluations filter optical depth estimates equidistant sampling comb. Combined use online control variate built from sampled mean estimate, resulting effectively performs marching most time while rarely-sampled higher terms correct bias.