A new algorithm for the estimation of 3-D projective invariants from uncalibrated images is proposed. It solves the uniqueness of the estimation of 3-D projective invariants of six points from uncalibrated images by introducing a 4th image. It also estimates the 3-D projective invariants by using two linear independent solutions of linear equations and is simpler than other algorithms which estimate the 3-D projective invariants by solving the nonlinear equations.