融合前景先验和背景先验的显著性目标检测

Fusion Foreground Prior and Background Prior Saliency Object Detection

现如今显著性目标检测得到越来越多学者的关注,同时已经成为许多计算机视觉任务中的重要使用工具。提出一种基于前景先验和背景先验结合的显著性目标检测框架。首先,从一幅图像中提取一系列的有效前景种子点和背景种子点,然后分别用于显著图计算。其次,将依靠于前景种子点和背景种子点各自生成的显著图结合。最后,利用测地线距离对显著区域进行进一步的增强。提供一种优选种子点的方式,从而得到更加稳定的种子点,在提高显著性精度方面有比较大的作用。

Nowadays,significance target detection has attracted more and more attention from scholars,and has become an important tool in many computer vision tasks.Proposes a framework of significance target detection based on the combination of foreground prior and background prior.First,extracts a series of effective foreground seed points and background seed points from an image,and then uses them respectively for the calculation of significance graph.Secondly,it will depend on the combination of significant graphs generated by the foreground seed points and the background seed points respectively.Finally,the geodesic distance is used to further enhance the significant region.Provides a way to optimize seed points,so as to obtain more stable seed points,which plays a greater role in improving the precision of significance.