Co-saliency-regularized correlation filter for object tracking
作者：
Highlights：
• The three-dimensional spatial and semantic saliency regularization is exploited to obtain the regularized weights with the spatial saliency information. The heterogeneous saliency fusion method accurately represent spatial co-saliency regions.
• Temporal saliency regularization weights are exploited to obtain object change information between frames and limit the change rate of the response map, as well as effectively and efficiently suppress abnormal mutations in the response map.
• Object tracking standard benchmarks including OTB-2015, UAV123, LaSOT and GOT-10k are used to evaluate the performance of the proposed tracker. Regarding tracking robustness and precision, the proposed tracker is compared with a variety of SOTA trackers.
摘要
•The three-dimensional spatial and semantic saliency regularization is exploited to obtain the regularized weights with the spatial saliency information. The heterogeneous saliency fusion method accurately represent spatial co-saliency regions.•Temporal saliency regularization weights are exploited to obtain object change information between frames and limit the change rate of the response map, as well as effectively and efficiently suppress abnormal mutations in the response map.•Object tracking standard benchmarks including OTB-2015, UAV123, LaSOT and GOT-10k are used to evaluate the performance of the proposed tracker. Regarding tracking robustness and precision, the proposed tracker is compared with a variety of SOTA trackers.
论文关键词：Correlation filter,Object tracking,Co-saliency map,Co-saliency regularization,ADMM optimization
论文评审过程：Received 10 May 2021, Revised 17 August 2021, Accepted 27 January 2022, Available online 2 February 2022, Version of Record 11 February 2022.
论文官网地址：https://doi.org/10.1016/j.image.2022.116655
