The Improved Denoising Method Research Of Remote Sensing Image

Today, Remote sensing images are often corrupted with noise during acquisitionand transmission. The reduction in the corrupting noise is very important, since thenoise deteriorates the quality of an image and makes the tasks such as compressionand segmentations difficult. In order to obtain the clear and exact image information,the research of image denoising algorithms is still one of the most fundamental,widely studied, and largely unsolved problems in computer vision and imageprocessing.In image denoising, a compromise has to be found between noise suppressionand the preservation of the important image edge and detail feathers.In this paper, the development of image denoising and some new algorithms inrecent years in the world were introduced firstly. The basic theories and classicmethods of image denoising were introduced in the second chapter, which focus onwavelet transform that have proved to be powerful tools for image denoising, the keypoint of wavelet transform is that small wavelet coefficients are more likely due tonoise, and large coefficients due to important signal features. Based on the variousstatistical models of wavelet coefficients, many wavelet shrinkage methods have beenproposed. In third chapter, the image denoising method is mainly discussed intransform domain, Combining a novel2-D signal processing theory calledNonsubsampled Contourlet Transform(NSCT), which was constructed to achievebetter multi-resolution, multi-direction, shift-invariant, the Image Denoising Based onthe Improved Fuzzy-Shrink Algorithm was proposed, It is all known that noise isuncorrelated in the NSCT domain, With respect to this principle, we use a fuzzyfeature for single channel image denoising to enhance image information in waveletsub-bands and then using a fuzzy membership function to shrink waveletcoefficients.Accordingly, This feature space distinguishes between important coefficients, which belong to image discontinuity and noisy coefficient. In chapterfour, the image denoising algorithm is mainly discussed in spatial domain.a methodcalled BM3D Image Denoising with Shape-Adaptive Principal Component Analysiswas proposed, which solve the problem that traditional BM3D algorithm has the lessability of sparsely to represent the true-image data, the proposed method improvedBM3D in two aspects, first, employing image patches (neighborhoods) which canhave data-adaptive shape. Second proposing PCA on these adaptive-shapeneighborhoods as part of the employed3-D transform. In chapter five, a new methodcalled Using K-SVD algorithm for improving performance of Bayes-shrink imagedenoising techniques was proposed, which solved the problem that the final noisereduced image has limited improvement in existing wavelet thresholding methodsbecause they do not remove noise in the approximation band, therefore, the K-SVDtechnique for noise reduction is applied on the approximation band to alleviate thedeficiency of the existing wavelet thresholding methods. At last, the simulationexperimental results show that the proposed methods can effectively reduce Gaussnoise in remote sensing image and preserve the image edges, the proposed methods isobviously superior both in vision and PSNR.