Abstract Intensity-ratio based phosphor thermometry is a well-established technique to obtain surface temperature measurements, which however suffers from important pixel-to-pixel noise, mainly due to the signal-dependent photon shot noise. In order to enhance signal-to-noise ratio levels, spatial filtering is recognized as a common post-processing routine, with the implication of strongly alleviating the spatial resolution (SR) of measurements. In this study, a dataset of phosphorescence images using Mg4FGeO6:Mn4+ thermographic particles and an ICCD camera is constituted. Various spatial filtering strategies are applied to these images, such as software pixel binning, moving average and Gaussian filters, in order to estimate their relative performances in terms of SR and temperature uncertainty. In addition, a Fourier space low-pass Butterworth filter is benchmarked against these common filters. Results show that the pixel binning strategy provides a limited improvement in the temperature uncertainty when compared to the loss in SR. Conversely, the Gaussian and moving average filters are found to effectively enhance the temperature uncertainty, though the 5th-order Butterworth filter is more selective, by providing an excellent mitigation of high-frequency noise with a minor attenuation of low-frequency information. Eventually, a joint spatial-spectral filtering strategy is investigated, which however does not present significant advantages compared to a sole filtering strategy.