Font Size: a A A

Research In SAR Speckle Suppression Methods

Posted on:2008-12-27Degree:MasterType:Thesis
Country:ChinaCandidate:G H HuFull Text:PDF
GTID:2178360215467548Subject:Communication and Information System
Abstract/Summary:PDF Full Text Request
Synthetic Aperture Radar (SAR) is a kind of high resolution imaging radar, butspeckle exist in SAR image because of the coherent processing of the scattered signalsin SAR imagery. Speckle noise is main obstacle to the use of SAR image, sorestraining speckle noise is necessary to refine SAR images for further application.The theme of this thesis is to study techniques of SAR speckle suppression, with themain purpose to both restrain the speckle and preserve valuable details of SAR image.The basic concepts and principles of SAR speckle is presented first, and sometraditional spatial filters for the speckle removal is compared briefly, then the maincategories of despeckling methods in wavelet domain and related advantages arepresented. After illuminating characters of inter- and intrascale dependencies amongwavelet coefficients, this article presents the denoising methods base on dependenciesin wavelet domain, in which the geometrical bayesian algorithm is thoroughlyexplained. Base on analyses in wavelet domain, a new mended geometrical bayesianalgorithm is applied to the steerable pyramid, and experiment proves that the newmended algorithm based on steerable pyramid achieves better speckle suppressioneffect. Moreover, through experiments of traditional spatial filters and waveletdenoise algorithms, a compare on despeckling performances of such specklesuppression methods is maded, and the compare conclusion is a guide on practicalapplication.Smoothing the speckle and preserving the image detail texture is conflicting inthe existing speckle suppression algorithms, and fusion is a helpful approach toresolve this conflict, so a new despeckling algorithm based on adaptive block combinemethod is proposed.In this fusion algorithm, the SAR image is segmented and anadaptive block is used to combine denoised images by different existed algorithms.Experiment results show that the fusion process can achieve the advantage of differentalgorithms and therefore it has an explicit performance to traditional algorithms.Because this fusion algorithm is succinct and effective, it has many advantages inpractical application.
Keywords/Search Tags:SAR, Speckle suppression, Wavelet transform, Multi-resolution analysis, Steerable pyramid, Geometrical bayesian estimation, Adaptive combined method
PDF Full Text Request
Related items