摘要
使用颜色、形状、纹理等特征的基于内容的图像检索技术,将图像看作向量空间中的点,通过计算两点之间的某种距离来衡量图像间的相似度,然而在提取图像特征时相同类型的图像会出现不一致的特征,极大地影响了检索算法的准确率。针对该问题,提出一种稀疏低秩描述的多特征图像检索方法。通过对图像集的稀疏低秩描述,保持了相同类别特征的全局结构,同时也降低了对于局部噪声的敏感度,增强了检索算法的鲁棒性。在Corel图像集上的检索实验结果表明,该方法较已有的基于内容的图像检索方法有更好的检索效果。
The content based image retrieval method extracts the color,textural,shape features of images,which can be represented in the feature space,with similarities among them obtained by some distance between feature vectors.Its accuracy critically depends on the feature vectors.However,images in same class will have different features.This paper presented an image retrieval method based on sparse low-rank representation.After the low-rank components of each set was recovered,both the global mixture of subspaces structure and the locally linear structure of the features were captured.The experimental results show that the method not only has a strong robustness to the unstablefeatures,but also has a good retrieval performance.
出处
《计算机科学》
CSCD
北大核心
2014年第3期302-305,共4页
Computer Science
基金
国家自然科学基金(61103070)
国家科技支撑计划课题(2012BAF10B12)资助
关键词
基于内容的图像检索
稀疏低秩描述
特征提取
Content based imageretrieval
Sparse low-rank representation
Feature extraction
