As prerequisite and foundation, the study of infrared imaging characterization for different ground scene in different time under different meteorological condition is the key of infrared image registration and ATR. In order to find out the proper time and optimum scene which is suitable for image registration, a great deal of aerial image sequences is analyzed in different season and different time at same region. Furthermore, a new metrics is put forward to measure the variation of IR scene with some statistical parameters, such as mean, standard deviation, POE. The metrics is designed to describe different scene properties quantitatively so as to evaluate the stability of scene MWIR imaging .According to the result of MWIR and visible image registration with NPROD, the proposed method can be used to correctly select reference area and increase the probability of map matching.