This paper presents a method to correct temperatures obtained by infrared thermography using thermo couples as reference temperatures. The raw signal values of the infrared detector are compared to the temperatures measured with thermocouples. The relation between raw thermal value and body tempera ture includes three parameters which allow to determine a best-fit approximation between detector signal and body temperature. The resulting non-linear system of equations is solved numerically using the Levenberg-Marquardt procedure. These parameters are then used to calculate the corrected tempera tures of the body. 1 Introduction Infrared thermography is a powerful method for temperature measurement in turbomachinery research. At the 'Institut fOr Thermische Stromungsmaschinen' at the 'Universitat Karlsruhe' the infrared camera has been used in numerous investigations on heat transfer phenomena. So far, the results achieved were mainly used for qualitative reviews of complex surface temperature distributions. To allow for any quantitative analyses the data have to undergo some further cor rection or calibration procedure. This procedure has to account for many different effects (e.g. reflected radiation, gas radiation, translucency of the window) which have an impact on the thermography results. In most experimental set-ups matching real engine conditions, these effects are difficult to quantify. Therefore, any correction procedure on a pure theoretical basis is not very promising. Better accuracy can be achieved with a case specific calibration of the thermography system which is obtained experimentally. Various authors have proposed calibration procedures for thermography systems which rely on calibration experiments. Among them Koschel et al. [1] used an external system to calibrate a pyrometer which was used to measure the temperatures on the surface of turbine blades. In this way the influence of the observation angle and thE: distance between blade surface and detector on the IR-signal could be studied precisely. Nevertheless, the surface emissivity could not be determined accurately and reflection effects could not be simulated in the calibration setup. Carlomagno et al. [2] used a little black body to calibrate an IR-camera under real test conditions. The black body was placed in the test rig at the location at which the surface tem perature had to be determined later on. The calibration parameters obtained with the black body were then used in the surface temperature measurements with an additional correction for gray body. This method can only be applied if the test surface can be dealt as a gray body, i.e. its emissivity is independent of temperature, and if the test section allows for the insertion of a black body. Reflected radiation, which occurs in the case of any gray body, cannot be taken into account. Meyers at al. [3] performed temperature measurements on film cooled combustor walls. During the heating-up of the combustor, temperature data were recorded from both the infrared scanner and from the embedded thermocouples. These data were then paired and used for a best fit approximation for the calibration parameters of the camera. Thus, all tem perature dependent properties of the system are taken into account. Nevertheless thermal im age recording during the heating-up period leads to slightly shifted operating conditions be tween calibrating and testing. The calibration method presented in this paper uses a similar approach but thermocouple temperature and thermal image recording takes place at steady state and the 'best-fit' approximation includes three parameters instead of two for higher accu racy.