Imaging Spectrometer is an important tool for quantitative analysis and discrimination of ground objects. The images of Imaging Spectrometer contain random noises to varying degrees of severity in all channels. To optimize the use of the images of an Imaging Spectrometer and to improve the effectiveness and accuracy of discriminating ground objects according to spectral absorption features, removal of random noises of images is necessary. Based on the analysis of previous methods of removing random noises of images, this paper develops a new method of removing random noises from Imaging Spectrometer images based on wavelet analysis (RRNW). Experimental results show that the performance of this method is better than neighbor average and median filter in removing random noises and reserving richer fine textures and edge information.