In this paper, a region based multiscale image fusion based on discrete wavelet frame (DWF) is proposed to combine information from thermal and visible images. Because of the fusion result of thermal and visible image may then allow both detection and imperceptible localization of the target in thermal image with respect to the discernible background provided by visible image. Additionally, the objective is to decrease the misregistration problem of these images which is actually exists in practice. The concept of discrete wavelet frame, which leads to translation invariant, can solve this problem. In our experiments, the preprocessing procedure of these source images is approached. It includes enhancement and registration based on object matching. Experimental results show that the proposed fusion scheme provides more robust when mis-registration distortion exists for the source images.