3D SAR Image Background Separation Based on Seeded Region Growing

Abstract

The environmental interference and the noise in 3D synthetic aperture radar (SAR) image, considered as the background, are inescapable and ought to be eliminated. For 3D SAR image, there is a spatial separation of the target and the background. Therefore, it is possible to achieve the separation of the target and the background by image segmentation. Due to the complexity of the target shape and the large dynamic range of the SAR image, the background cannot be accurately separated through the amplitude information alone. In this paper, a method based on region growing is proposed to achieve 3D SAR image background separation utilizing the plural and the spatial information. The image enhancement matrix, constructed by the plural information of the SAR image, is implemented to improve the contrast of the image. The seeds are extracted by the weighted Otsu, and the weight is determined by the structure and amplitude information from the target. For the region growing, the growing process is achieved by the accumulation of the growing rate, which can suppress the growing of the noise. During the region growing, the stopping growing condition of each seed is independent and controlled by the seed threshold. The global threshold constrains the almost unrestricted growing of the seed whose amplitude is close to the noise amplitude. The results of the simulation and the experiments verify the performance of the proposed method is higher than that of the compared methods with three image evaluation criteria. Besides, we discuss the cost of computation and the influences of three important parameters to achieve a complete analysis.

Keywords

• Synthetic aperture radar (SAR)
• image segmentation
• region growing
• image denoising