基于稀疏表示的图像显著区域检测算法

摘要

Focusing on the problem of images salient region detection,we proposed a sparse representation-based salient region detection algorithm.First,the algorithm uses sparse coding to describe images feature.Then it calculates the visual saliency based on images sparse coding instead of directly processing raw image so as to improve the efficiency of computation.Finally,according to the computation result of visual saliency it segments salient regions.The proposed method was experimented on public test image datasets and the experiment was compared with some other current popular algorithms.Experimental results showed that this algorithm was correct and effective when applying in images salient region detection.%针对图像的显著区域检测问题，提出一种基于稀疏表示的显著区域检测算法。该算法首先利用稀疏编码对图像进行特征描述，然后根据图像的稀疏编码进行视觉显著性的计算，而不是对原始图像直接进行处理，提高计算的效率。最后，根据视觉显著性的计算结果，进行显著性区域分割。在公开的测试图像集上进行实验，并和目前几种流行的算法进行实验对比。实验结果表明，该算法用于图像的显著区域检测是正确有效的。