期刊名称：ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
印刷版ISSN：2194-9042
电子版ISSN：2194-9050
出版年度：2010
卷号：XXXVIII - Part 7A
页码：117-121
出版社：Copernicus Publications
摘要：The usual polarimetric speckle filters optimally combine the polarization channels into a single-channleimage or only restore theradiometric information. The phase differences and the polarization channel correlation coefficients are lost. Though simple boxcarfilter and extended Lee filter for covariance matrix can beused topreservepolarimetric information without changing the data form,the results are not satisfying enough. In this study, we proposed a new framework for polarimetric SAR filter in which phase andintensityinformationare filtered independentlybased on the independence of intensity and polarimetric information. Firstly,polarimetric covariance matrix is dividedinto the product of a scalar z and amatrixC0. The sum ofdiagonalelements inC0is 1 andzdenotesthe sum ofdiagonal elements incovariance matrixwhich is also called total power or SPAN.SPAN zandC0includeintensityor texture informationand polarimetric information respectively which both are relatively independent. Secondly,SPANimage is filtered as singlypolarized imagery, for example using Lee filter.To preservethe polarimetricinformation, the boxcar filteris appliedtoC0, which can be considered as averagingpixels in a moving window usingequal weightsfor each pixel. Also, thefilteredC0must be normalized by dividingeach element bythe sum of itsdiagonalelements. Finally, the filtered result is obtainedby multiplying the filtered SPANandfilteredC0.Experiment results show that the proposed filter is valid, and polarimetricinformation is well preserved. And compared with extended Lee filter, the performance of the proposed method is
关键词：Radar;SAR;Polarization;Imagery; Land cover; filter; speckle
