GVF Snake Based Target Contour Extraction in SAR Imagery

Abstract

Target contour provides valuable shape information that can be used in pose estimation, target recognition, etc., which makes contour extraction a hot research topic. Gradient vector flow (GVF) snake has been proved to be an efficient method for extracting target contour. Due to the existence of speckle noise, direct applying GVF snake for extracting target contour in synthetic aperture radar (SAR) imagery is not successful. Based on the characteristics of SAR imagery, a method of using GVF snake for extracting target contour in SAR imagery is proposed. The proposed method first computes a likelihood ratio image in which the contrast between target and background is enhanced, and then a line-drawing edge map is generated based on the computed likelihood ratio image, finally GVF snake extracts the target contour by using the generated edge map. The experimental results on the real SAR image have demonstrated the effectiveness of the proposed method.