摘要
提出一种图像分割的多特征融合加权稀疏子空间聚类方法。采用多种属性的特征能够更可靠地描述图像中不同物体的特性,提高分割的准确性和可靠性。定义了加权稀疏度量,即在1-范数中引入权重,权重与数据的相似度成反比,有利于迫使相似的数据尽可能参与到数据的自表示中,从而改善稀疏表示过稀疏并且不稳定的局限性。实验结果和客观指标表明,所提方法能有效地分割自然图像,获得的结果更加符合人类视觉感知。
A weighted-sparse subspace clustering method with multi-feature fusion is proposed for imagesegmentation. Integration of multiple features can reliably describe the characteristics of various objects in naturalimages, thus can improve the accuracy and reliability of segmentation. The weighted-sparse measure is definedby introducing weights in the 1-norm of vectors. The weight is inversely proportional to the similarity betweendata？ therefore the weighted 1-norm penalty on the linear representation coefficients tends to force similardata be involved while dissimilar data uninvolved in the linear representation of a datum. The resulted representationcan overcome the drawbacks of 1-norm penalty that the presentation coefficients are usually over sparseand not robust for highly correlated data. Experimental results and objective assessment indexes show that theproposed method can effectively segment natural images with good visual consistency.
出处
《系统工程与电子技术》
EI
CSCD
北大核心
2016年第9期2184-2191,共8页
Systems Engineering and Electronics
基金
国家自然科学基金(61472303
61271294)
中央高校基本科研业务费(NSIY21)资助课题
关键词
图像分割
多特征融合
子空间聚类
加权稀疏
image segmentation
multi-feature fusion
subspace clustering
weighted-sparse
