Context-based hierarchical saliency detection for mobile hotspot

Abstract

Embedded hotspot applications, especially advertisements, supporting by 3/4G technology are now rising rapidly. Defining hotspot by finding attention region using saliency is one of interesting. In this paper, we proposed context-based hierarchical saliency feature detection. To construct saliency map, two main features are considered as the context, local and global features. The areas that have distinctive color which obtains high saliency, and blurred areas which give low saliency are considered as local. Moreover, the frequency-occurring information is consider as background which is pressed is considered as global. In experiments, our proposed method comparing with conventional hierarchical saliency framework shows high the recall and F-measure value.