计算机工程

• 图形图像处理 • 上一篇    下一篇

基于拉普拉斯非负稀疏编码的图像分类

李钱钱,曹 国   

  1. (南京理工大学计算机科学与技术学院,南京 210094)
  • 收稿日期:2013-05-28 出版日期:2013-11-15 发布日期:2013-11-13
  • 作者简介:李钱钱(1989-),男,硕士研究生,主研方向:模式识别,图像处理;曹 国,副教授、博士
  • 基金项目:
    国家自然科学基金资助项目(61003108, 61371168)

Image Classification Based on Laplacian Non-negative Sparse Coding

LI Qian-qian, CAO Guo   

  1. (School of Computer Science and Technology, Nanjing University of Science & Technology, Nanjing 210094, China)
  • Received:2013-05-28 Online:2013-11-15 Published:2013-11-13

摘要: 针对复杂背景下的图像分类问题,结合非负稀疏编码和局部保持投影算法,提出一种拉普拉斯正则化非负稀疏编码算法。相比于已有的稀疏编码算法,该算法不仅能更好地模拟哺乳动物初级视觉系统主视皮层V1区简单细胞感受野的行为,同时也可使相似的特征经过编码后仍然相似,从而保证特征度量的一致性。将该算法与空间金字塔匹配模型相结合应用于图像分类,在多个图像数据库上的实验结果表明,该算法具有较高的分类精度。

关键词: 稀疏编码, 非负稀疏编码, 拉普拉斯非负稀疏编码, 空间金字塔匹配模型, 图像分类, 支持向量机

Abstract: Aiming at the problem of image classification with a complex background, this paper proposes a new image classification algorithm based on Laplacian regularization Non-negative Sparse Coding(LNNSC) algorithm. It is combined with the respective advantages of the Non-negative Sparse Coding(NNSC) and Locality Preserving Projecting(LPP) algorithm, compared with the sparse coding algorithm, the proposed algorithm not only can better simulate the main visual cortex V1 simple cell receptive field behavior of the mammal primary visual system, but also can make the non-negative sparse codes of them be similar to each other. This paper combines the proposed LNNSC with Spatial Pyramid Matching(SPM) model and further applies it to image classification. Experimental results verify that the proposed algorithm can achieve higher accuracy.

Key words: Sparse Coding(SC), Non-negative Sparse Coding(NNSC), Laplacian Non-negative Sparse Coding(LNNSC), Spatial Pyramid Matching(SPM) model, image classification, Support Vector Machine(SVM)

中图分类号: