The Hyper Spectral Imager (HSI) onboard the HJ-1A satellite has acquired a large quantity of image data since its launch in Sep. 2008. However, the effectiveness of those images is greatly limited because of the apparent stripe noise in the images acquired in partial bands of the HIS. According to the features of the image noise, a new method for removing the stripe noise by moving the windows based on a reference band is proposed. The result shows that this new noise removing algorithm can not only remove the stripe noise in the images effectively and improve the definition of the images remarkably, but also can retain the basic information of the raw images. The method has the advantages of fast processing and wide range of application. It is an ideal stripe noise removing method.