SAR Image Segmentation Based On GLCM Of Geometric Region And Region Map

SAR image segmentation is one of the basic and crucial technology in SAR imageprocessing and interpreting. The segmentation results have great influence on thesubsequent image processing. Because of the unique imaging mechanism, the regionsgathered in SAR images always have the regular shade gray scale characteristic and thisleads to the problem in existing SAR image segmentation methods that the same areamay be classed into different areas.The primary sketch model is the sparserepresentation of the image, with its sketch lines depicting the gray changes of theimage. SAR image segmentation method based on Primary Sketch Map and semanticinformation, proposed by LiuFang and Yuanjialin, analyzed the aggregation of neighborsketch lines, divided sketch lines to clustered lines and non-clustered lines according tostatistical distribution of lines’ aggregation. Then different area extracting operator weredesigned to extract more abstract area map on the sketch map and intermediate semanticlayer based on areas was constructed and segmentation result was gotten combined withthe traditional segmentation method.. Based on the mentioned method as well as thesketch lines and the GLCM, this paper proposes a SAR image segmentation methodbased on GLCM and Region Map of the geometry area, The main work of the paper isas follows:First, for the problems existing in the SAR image segmentation, we analyze theprimary sketch model theory and the GLCM theory. With the geometry information ofthe SAR image sketch lines gotten from the primary sketch model, we define thegeometry region of the sketch lines. We count up the GLCM of the geometry regionswith the sketch lines to show the different semantic information of the sketch lines. Wealso divide the sketch lines into two classes with the first class depicting the borders andthe gray changes, and the second depicting the shade gray changes. Based on theclassification and the clustering features of the lines, we extract the region map of thegathering region, border region, non-line region.Second, based on the Region Map extracted from the sketch, this paper presents aSAR image segmentation method based on Region Map. With the Region Map, themethod maps the original SAR image into gathering region, border region, non-lineregion. We segment the SAR image into many super-pixels using the watershedapproach. We use different methods to merge the super-pixels of different regions and then merge the results of the integration of the three regions. Finally, we cluster themerging results by AP methods to get the marks of different regions. Thus we realizethe SAR image segmentation. Experiments show that, the SAR image segmentationmethod proposed in this paper solves problem in existing SAR image segmentationmethods that the same area may be classed into different areas.