The Research For Remote Sensing Image Denoising Method Based On Ground Surface Spectrum Vector Space

There are some errors of remote sensing images acquired by ETM+remotesensors for earth observation, such as a high frequency component of atmosphericinterference, residual errors and other errors left early image processing of unknownorigin. So it requires denoising in practical applications. In order to improve imageclarity, scholars explore a variety of methods. The low-pass filtering is better, whichis widely used in remote sensing image processing software, but the filtering effect isnot satisfactory. It is mainly for internal weak feature remarkable clarity, the featureedge pasting weak significant weakening noise while weakening the signal. Denoisingmethod needs to be improved and the new method has yet to discover, so hope thatthe new method can weaken noise, enhance signal and output feature high precisionclearer images.Combined with a low-pass filtering principle, this paper proposes a feature basedon the spectral characteristics of the filter vector denoising MFS. It uses the Landsat-7ETM+after scaling surface reflectance spectra normalized generalized image to dothe filtering experiment. On the premise of keeping the image feature spectral features,edge features, texture features, the terrain factor, the BRDF factor of landmark and theratio factor of each groud object in the mixed pixel, it reduces noise and improves theaccuracy of pixel values and image clarity. MFS does not need DTM Data to adapt tochanges in terrain, the image denoising of mountains are equivalent to the flat areas,the image denoising of the dark areas are equivalent to the bright areas, panoramicimage denoising is consistent and balanced. MFS maintains the physical dimensionsof the original image unchanged, while enhancing feature signal denoising andimproving the image SNR for remote sensing image preprocessing.The paper is divided into five chapters, the first chapter introduces the researchbackground, research status, the classical algorithm remote sensing image denoisingand research background and significance. The second chapter introduces thereflection characteristics of the surface features described method, and then discussesin detail the theoretical and normalized spectral vector normalized spectral vectorsintroduced by the generalized vector normalized spectral theory, and finallyintroduces the algorithm MFS filter model. The third chapter introduces the filteringto achieve this article C#development language as well as according to the theory ofprogramming processes. The fourth chapter is a combination of the contents of the previous chapters to Landsat-7ETM+remote sensing image as the experimental data,ETM+bands denoising first, and currently used four kinds of filtering methods arecompared, MFS comparative advantage denoising effect obviously, is expected toreplace the existing remote sensing image de-noising method, there is a certain value.The fifth chapter is the conclusion of this paper, mainly on progress and problems inthis paper achieved and the outlook for further work.