Abstract: CBERS-02C satellite images are widely used in the fields of resource investigation, environment protection, and agricultural research, however, the tripe noise is a widespread problem which cannot be ignored. Moment Matching is often used by assuming that the mean and standard deviation of the data acquired by each detector are identical, but traditional methods may cause blurring and ringing artifacts due to the strong assumption of subdetectors viewing the same scene. This paper proposes an improved algorithm for Moment Matching to remove stripe noise in CBERS-02C images. A moving-window which can be resized adaptively is used to deal with a wide range satellite images and the size of the window will depend on the amount of information within it. At the same time, Gaussian weighted method is used to obtain the reference value of column averages and variances. Finally, the stripe noise will be removed with the moving-window from left to right. Experimental results show that the improved method, qualitatively and quantitatively compared with traditional Moment Matching methods, is more effective in eliminating strip noise and avoiding information distortion. Besides, the improved method is computationally efficient and automatic, which can meet the needs of auto production of massive satellite imagery.