Three-dimensional digital images are gaining more attention in pattern recognition field. Mostly literatures, however, only focus on theoretical framework of twodimensional moment invariants, that are only implemented on two-dimensional images. Consequently, it reduces the invariance flexibility to support three-dimensional objects. In this paper, we introduce three-dimensional scale invariants of Legendre moments. They are algebraically derived directly from Legendre polynomials. Simulated experiments using three-dimensional binary images are carried out to verify the validity of proposed invariance.