On the basis of voxel data for a plurality of voxels constituting a set of ultrasound volume data, a voxel group identifying unit 50 identifies, in said ultrasound volume data, one or more voxel groups formed by a plurality of voxels in which voxel data satisfy a condition of being linked. On the basis of the voxel data for a plurality of voxels corresponding to each voxel group to be displayed, from among the one or more identified voxel groups, an image forming unit 80 forms an ultrasound image in which the voxel groups to be displayed are indicated clearly in a selective manner. It is thus possible for a three-dimensional image to be formed in such a way that one part of the image, such as floating matter in the amniotic fluid, does not interfere with another part of the image, such as a fetus.