Image filtering preprocessing which is helpful for increasing the signal to noise ratio（SNR）

decreasing the intra-class spectral variability and spatially smoothing homogeneous areas on the image can prove very useful for further discrimination of ground objects

image segmentation and classification processing.In this paper

two nonlinear anisotropic diffusion filtering methods are presented and they are based on the multispectral anisotropic diffusion models proposed by Pope and Acton.We build a couple of new diffusion coefficients in partial derivative equation（PDE） based on Tukey’s biweight estimator error norm by recurring to the relationship between robust statistics and anisotropic diffusion incorporated with the nonlinear time-dependent cooling technique for gradient threshold.Our methods not only effectively remove the impulsive noise caused by sensors

but also preferably preserve important detailed edges and image quality in remotely sensed images.Experimental results are given to show that the improved methods have superiority capability over the multispectral anisotropic diffusion schemes proposed by Pope and Acton on visual judgment and quality statistical analysis and they are very ideal edge-preserving filtering methods.