微目标遥感图像的感兴趣区域提取新方法

Extraction of ROI for Small Object Based on Visual Attention Mechanisms toward Applica- tions on Remote Sense Image Processing

结合微目标遥感图像的灰度、轮廓、大小等信息特征,提出了一种基于视觉注意的微目标遥感图像感兴趣区域提取方法,首先采用形态学Top-Hat变换强化感兴趣区域和抑制背景,利用开运算实现去除噪声和虚假微目标;然后通过阈值迭代法初步分割出感兴趣区域;最后通过人工交互,结合形态学方法进行感兴趣区域提取,进而运用基于小波变换的多尺度边缘检测算法探测出感兴趣区域的边界。实验结果表明:该方法能快速有效地分割提取出微目标感兴趣区域。

Combined with the specialty of remote sense image such as gray, profile, size and so on, a new ROI (Region of Interest) extract method for small object is presented, which is based on visual attention mechanisms toward applications on remote sense im- age processing. First by Top- Hat transform of mathematical morphology, ROI is intensified, and the background of original image is suppressed. After that, the noise and illusive small objects are removed by open processing. Then ROI is partitioned by setting ...