METHOD OF AUTOMATED CORRECTION OF INSTRUMENT DISTORTIONS ON LANDSAT-7 MULTISPECTRAL SATELLITE IMAGES

Abstract

An alternative method of automated correction of instrumental distortions of the ETM + scanner on Landsat-7 multispectral satellite images is proposed. The method is based on the application of filtering in the field of spatial frequencies using fast Fourier transform and spectral masks. The parameters of the mask are determined by a known spatial period of interference or experimentally by the results of the filtration. To filter the area of spatial frequencies to be removed, a filtering mask is applied, consisting of pixels with zero values. Periodic components whose spatial period corresponds to the spatial frequencies filtered on the Fourier image will be removed from the filtered image. The quality of filtration is determined by the mask used. When using multispectral images, a filter mask based on the Fourier image of one of the spectral channels is applied to filter other channels. The results of visual analysis of the processed images showed a fairly high-quality correction and elimination of instrumental distortions of the ETM + scanner in comparison with other methods. The main advantages of the proposed method are the next: the ability to work without the use of metadata and masks of instrumental distortion; work with single-channel and multispectral images; ability to work with small fragments of the scene; minimum number of manual settings for processing procedures; possibility to work without additional pictures for other dates; high stability of the used algorithms when using images from different satellites; fairly good repeatability of the results on satellite images taken in different seasons and for different areas. Experimental testing of the proposed method on a large number of images also confirmed the good repeatability and high stability of the algorithms used. It is expected that the developed technology will also be successfully used to cor