This paper reports an algorithm for measuring the time-averaged skin friction vector field \(\overline{\pmb {\tau }}(\pmb {X})\) starting from time-resolved temperature maps, acquired by a functional coating of temperature-sensitive paint. The algorithm is applied to a large area around a wall-mounted cube, immersed in the turbulent boundary layer over a flat plate. The method adopts a relaxed version of the Taylor Hypothesis operating on time-resolved maps of temperature fluctuations \(T'\) measured on the slightly warmer bounding surface. The procedure extracts \({\overline{U}}_T(\pmb {X})\) , the celerity of displacement of \(T'\) , as the best approximation of the forecasting provided by the frozen turbulence assumption near the wall, where its rigorous application is inappropriate. The \(\overline{\pmb {\tau }}(\pmb {X})\) estimation is based on the hypothesis of a linear relationship between \({\overline{U}}_T(\pmb {X})\) and \({\overline{U}}_U(\pmb {X})\) , chained to the one between \({\overline{U}}_U(\pmb {X})\) and \({\overline{U}}_\tau (\pmb {X})\) . We assess the outcomes of the proposed algorithm against those derived by the 2D and 3D Lagrangian particle tracking (LPT) methodology ’Shake-The-Box’, whose advent has made available high-quality near-wall flow field information. Furthermore, data from high-density 2D time-resolved LPT allows exploring the suitability of the linear relationships chain between \({\overline{U}}_T(\pmb {X})\) and \({\overline{U}}_\tau (\pmb {X})\) in the proposed context.