基于全局颜色对比的显著性目标检测

为了能正确检测显著性图中的多个显著性目标, 提出了一种基于全局颜色对比的显著性目标检测算法。该算法首先提取图像的全局颜色对比度特征, 然后把显著性图和全局颜色对比度作为特征输入条件随机场框架中, 得到二值显著性掩模, 最后经区域描绘子计算得到包含显著性目标的最小外接矩形。在两种公开的数据集上的实验结果表明, 该算法在精度、召回率以及F-测度方面的表现优于现有其他几种算法, 在计算效率上也具有一定的优势。因此, 所提出的算法在检测效果上优于现有的显著性目标检测算法, 而且还能够检测到多个显著性目标。

Salient objects detection based on global color contrast

This paper proposed an algorithm based on global color contrast for detecting multiple salient objects in the saliency maps. First, it extracted the global color contrast feature from the image. Then, by combining the global color contrast feature with saliency map feature into the conditional random fields (CRF), it got binary saliency mask. Finally, it attained bounding boxes of salient objects by region-props method. Experimental results in the two public databases show that the performance of the algorithm was better than others in precision, recall, and F-measurement. The algorithm has an advantage in computational efficiency too. Therefore, the proposed approach is comparable to the state-of-the-art salient object detection algorithms and can detect multiple salient objects in the images.