SAR Image Despeckling Based On Nonlocal Means And The Application Research On Segmentation

Synthetic Aperture Radar(SAR) has the advantages of multiband, variable angle, strong penetrability over optical sensor. It's all-weather and all-time capability makes it applying widely in military and civil fields. However, because of the mechanism of SAR, speckle noise significantly degrades the image quality which increases great difficulties for the further processing of images classification, images segmentation, object detection and identification. Polarization SAR image is a kind of SAR image, which is also suffering from the effects of speckle noise. Despeckling has been one of the hotspots over the world. Image segmentation is one of the main research directions of image processing, too. As an important military targets, the airport detection and segmentation in SAR image has important research value. Existing segmentation methods are mostly according to the image characteristics, using a specific segmentation criterion for segmentation. Due to split sample are various, none of them can handle all cases, and each of them has its own limitations. Therefore, the study on it has never been stopped. The main work is as follows:(1) Pol-SAR image quality enhancement method based on bayesian nonlocal means. Combined with the framework of bayesian and nonlocal means filter, a weight calculation method is proposed and proved. Experimental results show that the weights calculated by proposed method are close to the data of real situations, and the filtering results have better tradeoff between speckle denoising and detail information preservation.(2) A segmentation method for airport SAR image based on local threshold of segmentation. Airports have obvious geometric features. Based on local threshold segmentation, using airport structure information, make the image of the blocks more reasonable, at the same time, intensity of illumination compensation function is introduced to deal with the intensity non-uniform problem in image. Tests on real airport SAR image show that proposed method in this paper can obtain the ideal segmentation results.