Graph-Guided Fusion Penalty Based Sparse Coding for Image Classification
文献类型：会议论文
| 作者 | Yang, Xiaoshan ; Zhang, Tianzhu; Xu, Changsheng; Xu CS(徐常胜)
|
| 出版日期 | 2013 |
| 会议日期 | 2013 |
| 会议地点 | 南京 |
| 关键词 | Image Classification Sparse Coding Smoothing Proximal Gradient |
| 英文摘要 |
In image classification, conventional sparse coding only encodes
local features independently. As a result, the similar local features
may be encoded into code vectors with large discrepancy. This
sensitiveness has became the bottleneck of the traditional sparse coding
based image classification methods. In this paper, we propose a novel
graph-guided fusion penalty based sparse coding method. To alleviate
the sensitiveness of the traditional sparse coding, our approach constrains
that the similar local features are encoded into similar code vectors. To
achieve this goal, we add the popular graph-guided fusion penalty term
into the traditional l1-regularized sparse coding formulation. Finally, we
adopt the multi-task form of the smoothing proximal gradient method
to solve our optimization problem efficiently. Experimental results on 3
benchmark datasets demonstrate the effectiveness of our improved sparse
coding method in image classification. |
| 会议录 | PCM
 |
| 源URL | [http://ir.ia.ac.cn/handle/173211/11762]  |
| 专题 | 自动化研究所_模式识别国家重点实验室_多媒体计算与图形学团队 |
| 通讯作者 | Xu CS(徐常胜) |
| 作者单位 | 中科院自动化研究所 |
| 推荐引用方式 GB/T 7714 | Yang, Xiaoshan,Zhang, Tianzhu,Xu, Changsheng,et al. Graph-Guided Fusion Penalty Based Sparse Coding for Image Classification[C]. 见:. 南京. 2013. |
入库方式： OAI收割
来源：自动化研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。