Abstract:

When traditional method of denoising from remote sensing image is used to remove image noise, it often causes the loss and blur of image details after denoising. In this paper, the two-dimensional EMD denoising theory is applied to the denoising of remote sensing images. An improved denoising algorithm for remote sensing images combined with two-dimensional EMD and adaptive Gaussian filtering is proposed. When denoising, the low-frequency information remains unchanged, only for the high-frequency information of the image. Different frequency IMF component maps after two-dimensional EMD decomposition use adaptively Gaussian filtering to denoise, so as to better denoise the noisy image. Through the comparative analysis of two groups of experiments shows that:the algorithm has larger peak signal to noise ratio, average gradient and structural similarity and smaller RMS error. And the edge detection results also show that when the noise is filtered out, the image after this algorithm denoising can be better retain the details and the edge profile information of the original image. All these show that the algorithm has better denoising effect.

Key words: remote sensing image, two-dimensional EMD, adaptive Gaussian filtering, denoising, comparative analysis