Temperature measurement in flames is a challenging problem. Recently, hyperspectral imaging has demonstrated to be able to provide accurate temperature maps in a standard flame. However, hyperspectral imagers are expensive instruments, and the data analysis is laborious. Thus, a more simple approach to temperature imaging would be advisable. Since important and systematic differences exist in the low-resolution spectra of flames as a function of their temperature and chemical composition, it is in principle possible to retrieve these parameters by means of multispectral imaging. In this work, a standard flame, whose temperature and CO2 concentration are known, is studied with an infrared camera in the MIR band (3 to 5 μm), provided with a six interference filter wheel. High- resolution emission spectra are calculated, using the HITEMP2010 database, as a function of flame temperature (T) and CO2 column density (QCO2 , measured in ppm·m), and integrated over the spectral transmittance profile of the selected interference filters. Measured radiances in each channel are compared to these simulated values and the absolute error is minimized at each pixel to retrieve values of T and Q, obtaining temperature and column density maps for the flame. Results are compared to the known values of the standard flame. First estimations of errors are found to be ΔT< 100 K and ΔQCO2 < 400 ppm·m for flames with T∼2200 K and QCO2 ∼3500 ppm·m. The possibility of reducing the number of filters and their effect on accuracy is studied.