航天遥感图像感兴趣区域的自动提取方法

感兴趣区域提取是航天遥感图像分析的重要前提。随着图像空间分辨率的提高,场景内显著目标以及背景变得愈加复杂。利用传统的特征提取技术将会耗费大量计算空间和时间。提出了基于改进视觉注意方法的感兴趣区域自动提取,在HSV空间将目标与背景在颜色和亮度上的差异作为显著特征,利用高斯金字塔和中心-周边求差算子计算图像的显著特征图,并对特征图进行归一化和线性融合,设计注意焦点的转移步骤,完成感兴趣区域的自动提取。通过仿真和实验可以看出,本方法能有效地实现航天遥感图像感兴趣区域的自动提取。

Method for Automatic Extraction of Region of Interest in Space Remote Sensing Images

Region of interest extraction is an important prerequisite in space remote sensing image analysis. As the spatial resolution of images increases, the context of significant targets and backgrounds within scenes become increasingly complex. Traditional feature extraction techniques may cost a great amount of calculation space and time. A region of interest automatic extraction method is proposed on the basis of improved visual attention. The color and brightness differences between targets and the background are used as significant features in HSV space. The salient feature maps are calculated by using Gaussian pyramid and the center-around difference operator, and then are normalized and linearly integrated. Transfer steps of focus of attention are designed to complete region of interest automatic extraction. Simulation and experiment show that this method can effectively achieve automatic extraction of region of interest in space remote sensing images.