A multi-stage area saliency detection model

Detecting salient areas of an image is becoming more and more attractive due to its wide application in many areas, such as image compression, robot navigation. This paper proposes a Multi-Stage (M-S) saliency-based visual attention model to detect the salient area. The new computation framework is inspired by the hierarchical human visual pathway and composed of multi stages. The proposed model utilizes both local contrasts and global regional contrasts saliency information to acquire the saliency map and outperforms other models in detecting distinguishing area from other disturbing salient areas. Experiments show the M-S model balances well between fast speed and good performance.