Salient Object Detection via Quaternionic Local Ranking Binary Pattern and High-Level Priors
Fan Wang, Guohua Peng · 2019
We proposed a novel salient object detection based on quaternionic local ranking binary pattern and high-level priors. Firstly, we employed a edge-preserving smoothing algorithm to smooth the minor variations within textures. Secondly, we adopted a local descriptor called quaternionic local ranking binary pattern (QLRBP) for smooth image, and got three feature maps. Futhermore, we presented a saliency measurement method based on Shannon entropy and L2norm for the feature maps. Finally, we combined the two level priors with feature maps to generate three initial salieny maps, and we obtained final saliency map by the saliency measure method. The experimental results demonstrate that our method is effective and efficient.