基于结构标签学习的显著性目标检测

Salient object detection based on structured labels learning

提出了一种基于结构标签学习的显著性目标检测算法,将结构化学习方法应用到显著性目标检测中。首先从含有标记的图像中随机采集固定大小的矩形区域,并记录其结构标签;然后使用含结构标签的区域特征构建决策树集合;最后采用监督学习的方法对图像进行优化预测,得到最终的显著图。实验结果表明,本文方法能较准确地检测出图像库中图像的显著性区域,在数据库MSRA5000和BSD300的AUC值分别为0.891 8、0.705 2,说明本文方法具有较好的显著性检测效果。

This paper proposes a salient object detection method based on structured labels learning, applying a structured learning method to salient object detection.Firstly,we get a fixed rectangular region randomly from the local image which includes the labeling,and record the corresponding struc-tured labels.Then,a collection of decision trees is built by using the regional features which includes the structured labels.Finally,the final saliency map is captured by using the supervised learning ap-proach.Experiments show that our method can detect the salient objects accurately,and the AUC scores are 0.891 8 and 0.705 2 on the MSRA5000 and BSD300 datasets,the result shows that our method can achieve good effect in salient object detection.