Hyperfusion-Net: Hyper-densely reflective feature fusion for salient object detection

作者：

Highlights：

• The HyperFusion-Net is proposed to learn complementary reflective features in the fusion view.

• A hyper-dense fusion method is proposed to integrate the global and local multi-scale features.

• The proposed method can capture clear object boundaries and spatially consistent saliency.

• State-of-the-art performance on seven challenging large-scale saliency benchmarks is achieved.

摘要

•The HyperFusion-Net is proposed to learn complementary reflective features in the fusion view.•A hyper-dense fusion method is proposed to integrate the global and local multi-scale features.•The proposed method can capture clear object boundaries and spatially consistent saliency.•State-of-the-art performance on seven challenging large-scale saliency benchmarks is achieved.

论文关键词：Salient object detection,Image reflection separation,Multiple feature fusion,Convolutional Neural Network

论文评审过程：Received 29 August 2018, Revised 6 February 2019, Accepted 7 May 2019, Available online 8 May 2019, Version of Record 10 May 2019.

论文官网地址：https://doi.org/10.1016/j.patcog.2019.05.012