摘要
显著性检测是计算机视觉研究的一个重要问题。提出了一种由底向上的基于稀疏表示的显著性检测新算法。一般显著性检测主要包含两个部分,即图像特征提取和显著性度量。对于一幅给定的图像,首先利用独立成分分析(ICA)方法提取图像特征,然后用一个局部和全局模型对图像进行显著性度量。其中,利用一种低秩表示方法提取全局显著性,以及利用一种稀疏编码方法提取局部显著性。最后融合局部和全局显著图得到最终的显著图。在一个人眼关注数据库上与目前几种流行的方法进行了对比实验,实验结果显示所提出的方法能够得到更高的视觉关注预测准确率。
Saliency detection has been considered to be an important issue in many computer vision tasks. We proposed a novel bottom-up saliency detection method based on sparse representation. Saliency detection includes two elements: image representation and saliency measurement. The two elements used in our method are both biological plausible and accurate. For an input image,we first used ICA algorithm to learn a set of basis functions, then the image could be represented by the set of basis functions. Next, we used a global and local saliency framework to measure the saliency respectively, and combined the two results to obtain the final saliency. The global saliency is obtained through Low-Rank Representation(LRR), and the local saliency is obtained through a sparse coding scheme. We compared our method with six state-of-the-art methods on two popular human eye fixation datasets. The experimental results indicate the accuracy of the proposed method to predict the human eye fixations is higher.
出处
《计算机科学》
CSCD
北大核心
2014年第10期80-83,116,共5页
Computer Science
基金
国家自然科学基金(60972124)
863项目(2009AA01Z321)
973计划子课题(2010CB327902)
高等学校博士学科点专项科研基金(20110201110012)资助
关键词
显著性检测
稀疏表示
低秩表示
稀疏编码
Saliency detection, Sparse representation, Low-rank representation, Sparse coding