多孔小波和非下采样滤波器组去除遥感图像的多种噪声

Remote Sensing Image Denoising by àtrous Wavelet and Nonsubsampled Directional Filter Bank

结合àtrous小波变换和非下采样轮廓波变换的优点,提出一种基于àtrous-非下采样轮廓波变换的遥感图像去噪方法.该方法用非抽取离散小波变换的àtrous算法对图像进行多尺度分解,然后用非下采样的多方向滤波器组对得到的细节分量进行多方向分解.对含有多种噪声的遥感图像,àtrous-非下采样轮廓波变换将图像中不同种类的噪声分解到不同的小波系数分量中,使得可以根据噪声特性选择最合适的去噪方法,比用一种方法去除所有类型的噪声更科学且去噪效果更好.

This paper proposed a method of denoising the remote sensing image based on àtrous-nonsubsampled contourlet transform.The method uses the àtrous wavelet—an undecimated discrete wavelet transform algorithm to decompose the image into two parts possessing approximate part and detail parts,which are the same size as the original image.Then the nonsubsampled directional filter bank is employed to decompose the detail parts into directional subbands.The different kinds of noises of the remote sensing image can be decomposed into the wavelet coefficients in different scale and directions,with which the best method can be chosen based on the characteristics of the different noises.It is more scientific and more effective than just using one method for all kinds of noises in the past.It is proved that the method proposed in the paper is more useful in removing the noise of the image,reserving richer fine textures and edge information than other traditional filtering methods.