The temperature of a heated foil, exposed to a turbulent flow, changes continuously in time due to the fluctuation of the convective heat transfer coefficient. Measuring with enough temporal resolution the foil temperature map and solving the unsteady energy balance of the foil can allow restoring the instantaneous value of the convective heat transfer coefficient. Nevertheless, the high characteristic frequencies of turbulent flows (especially in air) result in very small temperature fluctuations, which are typically hard to measure with an infrared camera. This work presents a novel filtering approach based on Proper Orthogonal Decomposition which allows to dramatically reduce the measurement random noise, thus enabling the detection of small temperature fluctuations and, as a consequence, to improve the estimate of instantaneous distributions of the convective heat transfer coefficient. The proposed methodology is tested on synthetic jets impinging on a foil, actuated with a loudspeaker at various actuation frequencies. The measurement capabilities of the present technique are assessed through comparison with measurements obtained from raw data and phase averaged measurements. (C) 2016 Elsevier Ltd. All rights reserved.