We show a method of realizing object tracking and image synthesis in the dark in which both target motion and a reference image are estimated. In iterative calculation, a broader search is performed by calculating differences after applying a strong low-pass filter to input images. As a result, we realized object tracking and image synthesis from simulated video images with an SNR of up to -6dB and real video images captured in a dark environment with an illuminance of less than 0.01 lx at the subject surface.