Algorithms using central projection all need to parameteriz e the obtained closed curves,which will increase the amount of calculation.Besid es,the accuracy of parameterization affects the effect of feature extraction.To this end,a central projection descriptor without paramet erization is proposed in this paper,which is equivalent to converting the image into a closed curve by the central projection transformation,and then whitening the region enclosed by the closed curve.The whitening transformation eliminates the slant,transforming the affine transformation into rotation and scale only,and finally the Fourier descriptor is used to extract the invariant features. Unlike the existing algorithms,the whitening transformation here is given in th e form of formulas and has no parameterization.The experimental results show that the classification accuracy of the proposed descriptor can achieve almost 100% in the absence of noise.In addition,the proposed descriptor is superior t o the traditional MSA and AMIs algorithms in anti-noise performance.This validates the effectiveness of the proposed alg orithm.