Nonlocal center–surround reconstruction-based bottom-up saliency estimation
作者：
Highlights：
• A nonlocal reconstruction-based bottom-up saliency estimation method is proposed.
• Nonlocal image processing is employed to provide a better treatment of textures.
• Sparsity pursuit is very effective for removing duplicated similar image regions.
• Reconstruction residual is inherently coherent with the bottom-up saliency.
• Our method is stable under three public datasets and consistent with human fixations.
摘要
Highlights•A nonlocal reconstruction-based bottom-up saliency estimation method is proposed.•Nonlocal image processing is employed to provide a better treatment of textures.•Sparsity pursuit is very effective for removing duplicated similar image regions.•Reconstruction residual is inherently coherent with the bottom-up saliency.•Our method is stable under three public datasets and consistent with human fixations.
论文关键词：Saliency,Compressed sensing,Nonlocal means,Exemplar-based image processing
论文评审过程：Received 18 September 2013, Revised 17 May 2014, Accepted 2 October 2014, Available online 23 October 2014.
论文官网地址：https://doi.org/10.1016/j.patcog.2014.10.007
