Multilevel Interactive Reverse-Guided Network for Salient Object Detection in Optical Remote Sensing Images

Salient object detection in optical remote sensing images (ORSI-SOD) is a challenging and relatively novel research field. While many convolutional neural networks have been employed for SOD in natural scenes, they often overlook the unique properties of ORSI. These images typically lack clear edges and sharp details, posing challenges in capturing essential features. To address these issues, we propose a new method for ORSI-SOD called the multilevel interactive reverse guided network, which coordinates high- and low-level features in a synchronized manner. The model consists of two key components: 1) global-pixel coordination module (GPC) and 2) the multilevel feature interactive module (MFI), which facilitate coordination between the encoder and decoder. Specifically, GPC extracts low-level features and fully mobilizes them to refine detection, while MFI captures high-level features and enhances the representation of salient regions through multilevel interaction. We present a multiscale receptive field module, which strengthens feature representation through the hierarchical combination of various unbalance dilated convolutions, while reducing adverse grid effects and enhancing inter-pixel continuity. In addition, to promote feature interaction, a reverse-guided module is introduced to compensate for the insufficient information in shallow features by leveraging deep features more effectively. Extensive experiments on two public datasets employing eight evaluation metrics demonstrate that MIRGNet outperforms 30 state-of-the-art methods.