The standard analysis of heat-camera measurements includes emitted as well as reflected radiation, but assumes the same emittance for these two contributions. This is not valid if the sample emittance has spectral structure, i.e., is nongray, in the working range of the detector and near the maximum of the Planck function. The temperature dependence of the averaged emittance is shown to cause a significant error in the emittance value. Two algorithms are derived, based on linearising the heat-camera emittance values as a function of sample temperature for correction of this error. They are demonstrated to be effective on a spectral step model as well as for experimental data. BeO has a steep emittance edge in the long wave atmospheric window, similar to the step model. As for the model calculations, the conventional heat-camera emittance was significantly different from the corresponding integrated average based on spectral, reflectance data. Heat-camera measurements are compared with integrated spectral reflectance of ceramic beryllium oxide to illustrate the corrections. The two correction algorithms improve the agreement significantly, and it is proposed that they should be used when the heat-camera emittance exhibits systematic temperature variation.