Saliency Region Detection via Local and Global Features

Saliency detection plays an important role in many applications such as image segmentation, retrieval, editor and object recognition. Salience detection methods mainly divide into two categories: local contrast method and global contrast method. AC method is one of the typically local contrast methods, and HC method is a classical method base on global contrast. In this paper, we proposed an approach combined with AC and HC method, it not only considers local features but global information. After synthesizing the salience map of the AC and HC methods, firstly, according to the characteristics of human visual system, we pay attention to the salience center of an image, and add a weight value to each pixel, the closer to the center, the higher the weight. Then, we enhance the contrast of the image, receding darker part and increasing bright one. To confirm the proposed approach's effectiveness, we compare AC and HC method with the proposed one. The experimental results show that the new method generates a better salience map.