中文说明：我们提出了一种新型的凸极效应 — — 上下文感知的显著性— — 旨在检测代表图像区域现场。这一定义不同于以前的定义其目标是确定固定点或检测占主导地位的对象。根据我们的显著性定义，我们提出了一种检测算法的基础在心理学的文学作品中观察到的四项原则。的拟议方法的好处被评价在两个应用程序只是占主导地位的对象的上下文中在哪里对象本身一样重要。在图像重定向我们证明使用我们的显著性可防止扭曲在重要的地区。在总结我们表明我们凸极有助于生产具有紧凑、 有吸引力和信息摘要。
English Description：
We propose a novel saliency effect, context aware saliency, to detect the scene of representative image regions. This definition is different from the previous definition, its goal is to identify fixed points or detect dominant objects. According to our definition of saliency, we propose a detection algorithm based on four principles observed in psychological literature. The benefits of the proposed approach are evaluated in the context that both applications are just dominant objects, where the object itself is just as important. In image redirection, we demonstrate that using our saliency can prevent distortions in important areas. In summary, we show that our salient pole helps to produce a compact, attractive and informative summary.