Saliency Detection Based on Dark Channel Prior and Foreground Saliency Probability
Jing Li, Xiaogen Zhou, Haonan Zheng, Qinquan Gao, Tong Tong · 2020
Great achievements have been made in resent saliency detection approaches. However, it is still challenging to detect accurate salient regions using these approaches when an object closely touches the image boundaries. To address the above problem, in this paper, we propose a novel model for saliency detection based on the dark channel and foreground saliency probability. First, we construct a linear combination image called color space volume based on the LAB color space, which can greatly highlight salient regions, while suppressing background regions. After that, a novel fusion algorithm is proposed to obtain a robust and uniform salient image based on the foreground saliency probability and weighted saliency probability map. Finally, experimental results on two large benchmarks demonstrate that the proposed method has achieved better performance than several state-of-the-art methods in terms of precision, F-measure, mean absolute error, and recall.