高光谱遥感图像频域自适应同态滤波薄云去除

HYPER-SPECTRUM IMAGE CLOUDS REMOVAL BY A NEW SELF-ADAPTIVE HOMOMORPHIC FILTER IN FREQUENCY DOMAIN

针对高光谱遥感图像波段多,各波段受云层影响程度不一致,云层本身不均匀等特点,提出了一种频域自适应同态滤波方法。该方法首先检测某个图像云层的范围和对应厚度,根据云层处于低频特点,利用变差函数确定滤波窗口,把云层按照空间邻域大小转换为傅立叶空间,根据云层厚薄调整高通滤波截止频率,然后反变换到空间域。该方法可解决频域空间同态滤波方法破坏无云区的信息,克服空间域滤波方法效果不佳,难以控制和解释的缺陷。结果表明,该方法在保持无云区信息不破坏前提下,较好地改善了不同厚薄云层的质量,计算机内存开销很小,对大数据处理速度较快。

It is essential to remove the clouds as it can interfere the image interpretation. but the traditional method of clouds removal, however, does not take into account the difference of clouds effect on different bands and the clouds changed itself, and takes so large memory to process the image that can not be accomplished sometimes when the image is too large. A new self-adaptive method is presented in this paper to overcome those disadvantages. Firstly, an arithmetic to detect the clouds and its thickness is developed, and then variogram is used to determine the size of gliding window in which the Fourier transform was carried out to remove clouds. A self-adaptive high pass filter is induced according to the change of clouds, and before transforming back to spatial domain, a formula is designed to decrease the border effect caused by fourier transform. and the results show that the efficiency of the method is higher than that of the traditional one.