The study of hyperspectral remote sensing data noise filtering algorithm is the key to improving data analysis. In
this paper, to remove the stripe noise in spatial domain smooth filtering algorithm by row was used, while the wavelet
threshold denoising method was used to filter random noise in spectral domain. The former was tested on actual data and
got better results comparing with other moment matching methods. Not only was the stripe noise weakened well, but
also the consistency of mean value curve was retained. Through the spectrum domain wavelet de-noising,the image is
more smooth adopting soft threshold denoising method comparing with hard threshold denoising method, this proves
that the soft threshold denoising filters the good effect. Experimental results demonstrated that the quality of the image
was improved and the radiative feature was retained.