In this paper we firstly use a new method——numerical differentiation, which was developed in recent years to calculate gray gradient, and then realize the inverse cloud motion wind by regularization. At last, through simulation and practical experiments, we compared the wind inverse results between the algorithms with or without gray gradient information when the cloud images include perturbation. The experimental results show that the new algorithm with gradient information can reduce the influence of image disturbance effectively, and also increase the precision of cloud motion winds. We are in a position to find a new way to cloud motion wind inversion.