A fast computational method of the multi-scale autoconvolution(MSA)transform is proposed in this paper.In order to reduce the times of MSA transform

the method deduces the smallest benchmark transform size according to the fast Fourier transform theory

and replaces the different transform sizes of the same scale transform within the minimum range of MSA transform scale.Then

for reducing the computational complexity

this method reduces the MSA transform by using the MSA transform symmetry outside the range of MSA transform scale.Several experiments on the aspects of time efficiency and accuracy of eigenvalue using typical sample data are given.The results demonstrate that computation speed of the fast proposed computational method is three times faster than that of the original method while maintaining eigenvalue accuracy.