New scale invariants of Tchebichef moments are proposed. Using the falling factor and the first Stirling numbers,Tchebichef polynomials are represented with the linear combinations of power series,therefore,scale factors in moments of scaled image are separated. Then,new power series are transformed back to Tchebichef polynomials. Consequently,scaled invariants of Tchebichef moments can be expressed as linear combinations of original moments. They have lower the computational complexity and no iterative error accumulations. Experiments show that the proposed descriptors have better performance in extracting scale invariant features of a binary image.