Multiview diffusion-based affinity graph learning with good neighbourhoods for salient object detection

Salient object detection is a challenging task in computer vision and has been used to extract valuable information from many real scenarios. The graph-based detection approach has attracted extensive attention because of its high efficiency and stability. Nevertheless, most existing approaches utilize multiview features to construct graph models, resulting in poor performance in extreme scenes. In graph-based models, the graph structure and neighbourhoods play essential roles in salient object detection performance. In this paper, we propose a novel saliency detection approach via multiview diffusion-based affinity learning with good neighbourhoods. The proposed model includes three components: 1) multiview diffusion-based affinity learning to produce a local/global affinity matrix, 2) subspace clustering to choose good neighbourhoods, and 3) an unsupervised graph-based diffusion model to guide saliency detection. The uniqueness of our affinity graph model lies in exploring multiview handcrafted features to identify different underlying salient objects in extreme scenes. Extensive experiments on several standard databases validate the superior performance of the proposed model over other state-of-the-art methods. The experimental results demonstrate that our graph model with multiview handcrafted features is competitive with the outstanding graph models with multiview deep features.