[en] We propose a novel image restoration approach to eliminate the interference fringes in terahertz (THz) imaging. By making use of the information in the background of a THz image, inverse image processing is introduced under a priori knowledge about the degradation mechanism to suppress the interference. Based on a simulated interference image and the original THz image of a piece of paper coated with graphite powder, taken from our continuous THz imaging system, the interference fringes are successfully eliminated. The enhancement of the peak signal-to-noise ratio and structural similarity index is achieved, compared with the frequency domain filtering. Moreover, the inverse image processing is performed on a THz image of fresh rat brain tissue. It presents an excellent result with accurate information for the brain tissue after interference elimination. (paper)