As a typical issue

the noise analysis and elimination of remote sensing images have attracted considerable attention

and become an important research field for remote sensing image processing. Although traditional de-noising methods can eliminate noises to some extent

the edges and details of image are usually blurred while eliminating noise. In addition

the classical P-M model cannot effectively remove the Gaussian noise near strong edges and details

while ROF model tends to produce fake edges even jaggies in smooth region. This paper proposes a local adaptive mixture model to address the issue. According to texture characteristics of an image’s local region

our model can adaptively adjust a weighting function. By this way

our model exploits the advantages of the P-M model in smooth local region

and the advantages of the ROF model in the region with rich textures or edges. This facilitates our model to effectively eliminate Gaussian noise

and at the same time well protect the edge features and details of remote sensing image. Experimental results show that our model gains 3dB and 2dB higher SNR than the P-M model and the ROF model do separately.