We have developed a tomosynthesis-based radioactive seed localization method for prostate brachytherapy. In contrast to the projection image-based matching approach, our method does not involve explicit segmentation of seeds and can recover hidden seeds. Modified distance map images are computed from a limited number of x-ray projection images, and are backprojected to reconstuct a 3-D volume of interest. Candidate seed locations are extracted from the reconstructed volume and false positive seeds are eliminated by solving an optimal geometry coverage problem. The simulation results indicate that the implanted seed locations can be estimated from three or four images depending on the number of seeds if the pose of a C-arm is known. The algorithm was validated using phantom and clinical patient data.