Abstract: Two layers of composite filter is a common way to remove mixed Gaussian and impulse noise in the image. Such methods do not consider the interaction between the two different layers and the shortcoming on the choice of filter threshold, thus the residual mixed noise is unavoidable, and filtering process is going to smooth the edge details of the image, leads to image observation fuzzy. So this paper proposes a method of de-noising that has three layers combination filter. Adaptive Wiener filtering is added to the third layer based on the wavelet BayesShrink threshold de-noising and the adaptive median filtering de-noising, this take the advantage of Wiener filtering in processing the noise when the signal is high SNR (signal noise ratio). In order to reserve the edge details of image in the de-noising process, this paper also uses edge detection operator, therefore the details are retained and the de-noising result is more sharpening. Experimental results show that the proposing method of triple layer composite filter is better than two layers of filter in de-noising the mixed Gaussian and impulse noise of remote sensing image.