Research On Speckle Reduction Methods For SAR Images

Speckle in SAR images results in image quality reduction and influences the visual effect.It also causes difficulties in further processing and analyzing such as image segmentation and object detection.Therefore,despeckling is a significant step in pre-processing.This dissertation mainly focuses on speckle reduction methods based on non-local theory.The main work of this paper is as follows:1.A downsampled SAR-BM3 D despeckling approach for SAR images in high resolution is proposed.SAR-BM3 D is executed on small scale after downsampling,and edges will be compensated in each level when upsampling.This approach not only utilizes the good ability of feature preservation,but also improves performance of smoothing homogenous regions.Edge compensation improve the effect further.When it comes to high-resolution SAR images,the efficiency can be raised by senven to eight times,compared to original SAR-BM3 D.2.Under the framework of weighted maximum likelihood estimation(WMLE),a novel idea,in which similarity is measured by the statistics in homogeneous area.This idea is inspired by ideal estimation procedure,which makes weighting map in accordance with the real homogeneity.Experimental results show that the similarity measure based on homogenous statistic characteristic gives rotation invariance and much better weighting map.The proposed method outperforms the art-of-the-state despeckling algorithms with ideal homogeneous measure.Results of the proposed method is better than the ones of PPB algorithm when using the homogeneous measure mentioned in the paper.At last,two idea about how to calculate homogenous set are presented.