Abstract: Human vision system can assign processing resource efficiently by saliency detection of different interesting objects in the scene. The saliency detection method based on the visual attention mechanism is employed to simplify the scene analysis and the target interpretation of remote sensing images, which economizes processing resources. On the foundation of visual attention mechanism, a scale self-adaptive saliency detection method of SAR image is proposed. The local complexity metric and self-dissimilarity metric of multiple scales are utilized to compute the saliency metric. Moreover, the way of saliency scale determination has been designed. The saliency map has been built by combining saliency metric with the saliency scale, which is the last step of saliency detection. Experimental results show that the proposed method can detect the saliency of SAR image effectively, and that the proposed method is more reliable for SAR image scene analysis than other state-of-the-art saliency detection methods.

Key words: Synthetic Aperture Radar（SAR）, self-adaptive, saliency, saliency map

摘要： 人类视觉系统能够通过对场景中感兴趣的不同事物进行显著性检测，有效地配置处理资源。基于视觉注意机制的显著性检测方法能够简化遥感影像场景分析、目标解译的复杂程度，节省处理资源。以视觉注意机制为基础，提出了一种尺度自适应的SAR图像显著性检测方法，通过不同尺度下的局部复杂度和自差异性来度量图像的显著性测度，设计显著性尺度确定算法以及融合显著性尺度和显著性测度以生成显著图，完成显著性检测的流程。实验结果表明该方法能够有效应用于SAR图像显著性检测，较之其他主流显著区域检测算法更适用于SAR图像场景分析。

关键词: 合成孔径雷达, 自适应, 显著性, 显著图