Polarimetric SAR Image Segmentation Based On Noise Suppression

Synthetic Aperture Radar (SAR) is an active imaging sensor, with the advantages of all-time, all-weather, high resolution, and it has been used in many areas.With the development of Synthetic Aperture Radar technology, the spatial and temporal resolution that Po1SAR can obtain keep improving. Processing the SAR image data would be rather difficult due to the impact of speckle noise, so it is of great significance to analyze the obtained SAR images. Image segmentation is an important part of image interpretation. In this dissertation, the noise suppression Po1SAR image segmentation algorithm is proposed.The main contents and the main innovations of this dissertation are as follows:1 We propose SAR sea ice image segmentation algorithm with noise suppression. The main advantage of this method is using speckle noise filtering before the watershed. Then we consider the degree of difference between the regions to establish a more accurate spatial context model for accurate SAR image segmentation.2 We propose variable weights polarimetric SAR image segmentation algorithm with noise suppression. We consider the region adjacency graph and make full use of boundary information to obtain adaptive weight. It can describe the relationship between the feature model and spatial context model effectively. We can segment SAR sea ice image more accurately.By experiments for the full polarimetric SAR sea ice images acquired by RADARSAT-2 and SIR-C, the new approach is superior to several advanced ones, where it has higher segmentation accuracy.The proposed algorithm would be a meaningful progress towards an accurate segmentation and interpretation of Po1SAR images.