An intelligent hybrid measurement method, which is successfully developed for linear elastic deformation fields (Nishioka et al. Exp Mech 40:170-179, 2000), is investigated for steady-state temperature field. First, a variational principle is derived to minimize the errors and noises associated with experimental temperature measurements. This variational principle assures also the satisfaction of the governing steady-state heat conduction equation. On the basis of this variational principle, an intelligent hybrid experimental-numerical method is developed for the measurement of temperature field, and for the subsequent visualization of higher-order quantities such as the heat flux distribution. Furthermore, a concept is presented for selfrestoring heat generation that automatically restores temperature field against experimental errors and noises. The present intelligent hybrid method successfully demonstrates automatic detection and automatic elimination of experimental measurement errors, and smooth visualization of heat flux distribution.