EGNet is a novel edge guidance network designed for salient object detection. It addresses the issue of coarse object boundaries in existing fully convolutional neural networks (FCNs) by focusing on the complementarity between salient edge information and salient object information. The network consists of three main components: progressive salient object feature extraction, non-local salient edge feature extraction, and a one-to-one guidance module. The progressive feature extraction step extracts multi-resolution features, while the non-local feature extraction integrates local edge and global location information to obtain salient edge features. The one-to-one guidance module fuses these complementary features at various resolutions to enhance salient object localization and segmentation. Experimental results show that EGNet outperforms state-of-the-art methods on six widely used datasets without any pre-processing or post-processing. The method achieves the best performance under three evaluation metrics, demonstrating its effectiveness in accurately detecting salient object boundaries and improving overall detection accuracy. The network's design allows for better utilization of complementary information, leading to more accurate and coherent salient object detection results.EGNet is a novel edge guidance network designed for salient object detection. It addresses the issue of coarse object boundaries in existing fully convolutional neural networks (FCNs) by focusing on the complementarity between salient edge information and salient object information. The network consists of three main components: progressive salient object feature extraction, non-local salient edge feature extraction, and a one-to-one guidance module. The progressive feature extraction step extracts multi-resolution features, while the non-local feature extraction integrates local edge and global location information to obtain salient edge features. The one-to-one guidance module fuses these complementary features at various resolutions to enhance salient object localization and segmentation. Experimental results show that EGNet outperforms state-of-the-art methods on six widely used datasets without any pre-processing or post-processing. The method achieves the best performance under three evaluation metrics, demonstrating its effectiveness in accurately detecting salient object boundaries and improving overall detection accuracy. The network's design allows for better utilization of complementary information, leading to more accurate and coherent salient object detection results.