中文摘要：

现有的显著性检测算法主要是由数据驱动的，是通过底层特征来检测能吸引用户注意的所有显著性区域，而在具体的图像理解过程中，用户往往关注某一种或几种特定目标。文章提出了一种目标驱动的显著性检测算法，得到用户感兴趣的显著性区域。该算法是用图像中像素点与目标图像的平均色之间的差异来表示单个像素点的显著值。结果表明，目标驱动显著性检测算法可以快速地检测出与目标图像相似的显著目标。

英文摘要：

The existing saliency detection methods are mainly driven by data ,and detect all the salient regions which can attract users’ attention by low-level features .However ,the users always pay at-tention to one or several kind of targets in the specific image understanding process .In this paper ,a target-driven saliency detection method is proposed to get the salient regions that users are interested in .The pixel saliency is defined by using a pixel’s color difference from the average image color of the target image .The experimental results demonstrate that the proposed method can quickly detect the salient object that is similar to the target image .