Change Detection Based On Local Information Statistic In SAR Image

Change detection in SAR Image is a process that analyzes images of the samescene taken at different times in order to identify changes that may have occurredbetween the considered acquisition dates. SAR image change detection techniques havebeen used successfully in many applications, such as investigation of forest resources,studies on land use/land cover dynamic detection, assessment of environment disaster,arrangement of urban growth, and monigoring of national defense, etc.In this paper, we focus on the issues related to how to extract change informationeffectively from SAR images. The major works can be summarizied as follows:(1) A novel method of change detection based on neighborhood similarity andimage enhancement is proposed. The neighborhood similarity operator is used to get thedifferenc image, then, the difference image is processed by image enhancement, whichcan retain the details and reduce the speckle noise of the difference image. Experimentalresults show that the proposed method can effectively concentrate and extract changeinformation, which will improve the accuracy of change detection.(2) In order to address the effection of speckle noise and registration errors, thispaper presents a novel approach by combining gray levels based on both their spatialcloseness and their photometric similarity. The spatial closeness aims at reducing thespeckle noise in SAR images and the photometric similarity maximize the separabilitybetween changed and unchanged classes. The proposed approach is characterized by adifferent tradeoff between speckle reduction and preservation of geometrical details.Compared with the traditional ratio operator, log-ratio operator and mean-ratio operator,the experimental results show that the proposed approach is superior to the traditionalmethods, and produce better detection results.(3) Another novel method of change detection based on bilateral similarity andLaplace transform is proposed. First, the use of bilateral similarity constructs twodifferent images with different characteristics. Second, this paper adopts a Laplscianpyramidal decomposition method to analyze two difference images, and fuse thedecomposed images with the method based different methods. By doing the reversedLaplacian pyramidal transform in the next step, the final differenc image is got.Fanally,thresholding it based on the Kittler-Illingworth (KI) threshold selection criterion.Experiments confirm the effectiveness of the proposed technique.This paper was supported by the National Natural Science Foundation of China (No.60703107), the National High Technology Research and Development Program(863Program) of China (No.2009AA12Z210), the Program For New Century ExcellentTalents in University (No.NCET-08-0811), the Program for New Scientific andTechnological Star of Shaanxi Province (No.2010KJXX-03) and the FundamentalResearch Funds for the Central Universities (No.K50510020001).