Nested Network With Two-Stream Pyramid for Salient Object Detection in Optical Remote Sensing Images

Arising from the various object types and scales,diverse imaging orientations, and cluttered backgrounds inoptical remote sensing image (RSI), it is difficult to directlyextend the success of salient object detection for nature sceneimage to the optical RSI. In this paper, we propose an end-toend deep network called LV-Net based on the shape of networkarchitecture, which detects salient objects from optical RSIs ina purely data-driven fashion. The proposed LV-Net consists oftwo key modules, i.e., a two-stream pyramid module (L-shapedmodule) and an encoder–decoder module with nested connections(V-shaped module). Specifically, the L-shaped module extractsa set of complementary information hierarchically by using atwo-stream pyramid structure, which is beneficial to perceivingthe diverse scales and local details of salient objects. TheV-shaped module gradually integrates encoder detail featureswith decoder semantic features through nested connections,which aims at suppressing the cluttered backgrounds and highlighting the salient objects. In addition, we construct the firstpublicly available optical RSI data set for salient object detection,including 800 images with varying spatial resolutions, diversesaliency types, and pixel-wise ground truth. Experiments onthis benchmark data set demonstrate that the proposed methodoutperforms the state-of-the-art salient object detection methodsboth qualitatively and quantitatively.