Because of disturbance from inside or outside of sensors

there is much noise in the remote sensing image.People always filtrate image in order to improve its quality.Traditional filters are based on image’s space dimension.As a result

they usually blur the edges while smooth the image;or make image rough while enhance image’s edges contrarily. The adaptive filter in this articls

AFS（Adaptive Filter based on Spectrum）

uses the spectral information of multispectral or hyperspectral remote sensing images to smooth the even object

strengthen edges and keep the spectra of exceptional point

line orparcel of image. The key of AFS is how to complete the choosimg pixels in one filter mask based on pixels’spectral information. This method is also a kind of convolution operation to images

the difference between AFS and traditional smooth method is operating with pixels in filter mask selectively.We give some weight to valid pixels and give zero weight to invalid pixels in the filter mask

then use this filter to operate with image.