A Novel Spatial Information Model for Saliency Detection

作者：Gao, Hang; Li, Bo^*; Liu, Han

来源：14th International Symposium on Neural Networks (ISNN), 2017-06-21 To 2017-06-26.

DOI：10.1007/978-3-319-59081-3_26

• Saliency detection
• Spatial information
• Image processing

Saliency detection is one of the critical issues in computer vision. The location of saliency object is a widely used cue in the procedure of detecting saliency map, which is named as spatial information. Center prior and Harris-points are two popular cues of spatial information. In this paper, we propose a novel spatial information model which is different from center prior and Harris-points. In the proposed method, a cost function and a rectangle are used to seek the saliency object accurately, and to discriminate the saliency object from background effectively. The model can be used to optimize previous saliency detection approaches. In the experiment, the model is used to optimize five state-of-the-art approaches on two popular datasets ASD and ECCSD. The experiment results demonstrate the feasibility and validity of our method. And the performance of our method is better than center prior and Harris-points.

• 出版日期2017
• 单位华南理工大学