Abstract:Salient object detection has been widely used in image retrieval, image segmentation, pedestrian recognition and other fields. Current mainstream detection methods fuse multi-level feature information through short connection to add feature maps, which cannot accurately and effectively control the transmission of information flow. In addition, existing salient detection methods usually use single feature detection, which results in discontinuous and fuzzy boundary between the saliency object region and the background. A new salient object detection method based on multi-scale feature extraction and multi-level feature fusion was proposed in this paper. Firstly, the multi-scale context information was obtained by using the dilated convolution of different expansion rates to make up for the deficiencies caused by single feature detection. Secondly, a multi-level feature fusion module was designed, which fuses low-level feature, high-level feature and global context feature information for different distribution characteristics of them. It can not only restrain the transmission of noise, but also restore the spatial detail structure information of the saliency object effectively. At the same time, a concise attention module was constructed, which can effectively retain the channel information after feature map fusion. The F-score, mean absolute error, structural measurement, precision-recall rate curve and F-measure curve have been evaluated experimentally. Experiments on five public datasets show that compared with other thirteen mainstream detection methods, the proposed method in this paper achieves significant improvements in different evaluation indicators, of which the F-score, and structural measurement on four datasets are better than other methods. Meanwhile, the saliency map predicted by the proposed method in this paper has better continuity of edge contours and clearer details of spatial structure details.