Local Means and Local Standard Deviations method（LMLSD） is one of the most widely used methods for estimating the noise in remote sensing images.It employs local means and local standard deviations of small imaging blocks and can be used to evaluate images with additive noises.But this method is badly affected by land cover types

when the land cover types of the images vary greatly

estimation of noise can be invalid due to the effects of the heterogeneous blocks containing edges and texture features.In order to reduce the sensibility of LMLSD to land cover complexity

two improved methods were designed

one is to reduce the affection of the heterogeneous blocks

and the other is to pick up the Gaussian distribution that shows the characteristics of the heterogeneous blocks.We validated the two methods with two AVIRIS radiance images acquired in the same aerial experiment.In the validation the improved methods show distinctly enhanced robustness compared to the common LMLSD and the estimation of the noise is proved to be more accurate.