基于边缘效应和注意力机制的显著性检测

Saliency Detection Based on Boundary Effect and Attention Mechanism

本文重点研究了显著边缘信息与显著目标信息的互补性,提出了一种结合注意力机制的边缘效应网络。采用逐步融合的方法提取图像中具有显著性的局部边缘信息与全局位置信息,得到了显著的边缘特征和显著的对象特征,最后在不同分辨率下将边缘特征与对象特征耦合起来,通过注意力机制进行优化,进一步提高显著性区域的特征权重,从而得到最终的显著图。综合实验结果表明,该方法在不需要任何预处理和后处理的情况下,在5个常用数据集的性能优于现有的方法。

This paper focuses on the complementarity between salient boundary information and salient object information,and we propose a kind of boundary effects network combining attention mechanism.The method of gradual fusion is used to extract the significant local boundary information and global position information in the image,so the salient boundary features and salient object features are obtained.Finally,the boundary features and object features are coupled under different resolutions,and the feature weight of the salient regions are further improved through the optimization of attention mechanism,so as to obtain the final prediction.The experimental results show that the performance of this method is better than that of the existing method in five common datasets without any preprocessing and post-processing.