Infrared thermography is often used to record an image of the building envelope surface temperature, and thereby acquire qualitative information on its thermal insulation performance. Recently, a thermography method has evolved, which enables quantitative measurement of the 2-dimensional pattern of the density of heat flow rate (q) across the building element surface. However, based on previous estimates of its measurement uncertainty, the capacity of the thermography method to yield accurate results has been questioned. We present here an improved procedure for measurement of q, with an evaluation of measurement errors. The main improvement consists of the simultaneous measurement of surface temperature, surrounding radiative temperature, and air temperature, based on information included in one single thermal camera image. This arrangement allows for accurate measurements of small temperatures differences, and thereby reduced uncertainty in the measurement of q. The measurement bias was evaluated experimentally by a comparison of thermography results against a reference method. Under natural convective conditions, there was a 2.6 W m(-2) constant difference between the two methods. The measurement uncertainty u(q) was estimated as a function of q. Based on this, the lower limit of the measurement working range was determined to be 6 W m(-2), which corresponds to less than 10% relative uncertainty. In the case of forced convection, the thermography method yielded less reliable results. The reason for this was the sensitivity of the results to the choice of model for the convective heat transfer coefficient, and the difficulty to select the position for measurement of the wind speed, which is appropriate for this model. (C) 2014 Elsevier Ltd. All rights reserved.