Image restoration techniques are studied for Compton backscatter imaging as applied to identification of a land mine buried in soil. Mathematical methods are developed to restore images, which include artifacts due to photon noise, soil surface irregularity, and vertical motion of the imaging system. The image restoration is formulated as an inverse photon transport problem. The forward photon transport is modeled by using a two-collision response function. The inverse problem then is solved by applying an iterative minimization algorithm, resulting in an estimation of characteristic parameters of objects. Mathematical relations among detector responses are derived by experimentally analyzing the detector response characteristics when there are soil surface irregularity and vertical motion of the imaging system. These are used to remove the artifacts from the images. The method successfully restores the geometrical feature of the object under simulated battlefield imaging conditions.