Stereo digital image correlation (Stereo-DIC) is recurrent in photo-mechanics to measure kinematic fields which can be of high interest for instrumenting open-mould forming processes. Nevertheless, in the presence of pre-heating operations, as observed in the context of thermoforming processes, natural convective heat flows risk emerging and causing optical distortions in the recorded images. Consequently, this alters the precision of the measured full-fields of displacements. To address these challenges, this study proposes an experimental approach with two distinctive features. Firstly, it focuses on regenerating the heat haze effect at a laboratory scale within a partially opened vertical enclosure and without utilizing any filtering air flows. Secondly, the study quantifies the spatial and temporal variations of errors through statistical analyses of the differences between measurements obtained from quasi-static speckle translations and known imposed displacements. Experimental results indicate that the main cause of displacement errors is related to the 3D nature of the hot air turbulence caused by the natural convection phenomenon. This observation is supported by the detection of feather-shaped heat flows causing optical out-of-plane surface deviations. Furthermore, the study validates the possibility of obtaining time-dependent corrective functions for bias errors, which characterize the performance of the calibrated Stereo-DIC system in the presence of heat haze. Despite the limitation of extensive measurements required by the proposed approach, this study contributes to addressing the heat haze effect and constitutes a step towards extending the use of stereo-DIC for in-situ instrumentation of short-duration thermomechanical tests in the presence of heat haze.