基于低级特征图最优融合的显著性目标检测

Salient object detection based on optimally combining low-level features

针对低水平特征显著性目标检测算法在检测图像时不能检测到不同大小的目标,而且精确度较低的问题。提出一种新的算法,通过将颜色分布,方向对比度以及基于频率信息这3种特征运用条件随机场进行最优全值线性融合后,得到更精确的显著性目标。通过与10种经典的显著性目标检测算法进行的定量和定性的对比,实验结果表明,提出的算法不仅可以有效地检测到大、中、小显著性目标,而且检测的效果比其他算法精确度高。

Salient object detection algorithm based on the low levels in the image does not detect different targets, nor have high accuracy. A new algorithm is proposed--the three characteristics of color distribution, direction contrast and frequency information are used for the optimal linear fusion to get a more accurate salient object. After the quantitative and qualitative comparison with 10 kinds of classic salient object detection algorithm, the experimental results show that the proposed algorithm not only can effectively detect the large, medium and small salient objects, but also has higher precision than other else.