计算机工程与应用 ›› 2019, Vol. 55 ›› Issue (10): 141-145.DOI: 10.3778/j.issn.1002-8331.1802-0148

• 模式识别与人工智能 • 上一篇    下一篇

图像分类中基于分类矢量量化的视觉词袋模型

王  娇1,罗四维2,邹  琪2   

  1. 1.国家开放大学 计算机科学与技术学院,北京 100039
    2.北京交通大学 计算机与信息技术学院,北京 100044
  • 出版日期:2019-05-15 发布日期:2019-05-13

Bag-of-Visual-Words Model Based on Classified Vector Quantization and Its Application in Image Classification

WANG Jiao1, LUO Siwei2, ZOU Qi2   

  1. 1.Institute of Computer Science and Technology, The Open University of China, Beijing 100039, China
    2.Institute of Computer and Information Technology, Beijing Jiaotong University, Beijing 100044, China
  • Online:2019-05-15 Published:2019-05-13

摘要: 特征表示是图像识别和分类的基础,视觉词袋是一种图像的特征表示方法。分析现有视觉词典构建方法的不足,提出一种新的视觉词典构建方法。首先利用梯度方差把特征矢量分为光滑类和边缘类,然后分别针对不同类别的特征矢量进行视觉词典的构建,最后根据两类视觉词典生成视觉词袋。图像分类实验表明,提出的新方法能提高分类准确率。

关键词: 视觉词袋, 图像分类, 矢量量化, 特征表示

Abstract: Feature representation is the basis of image recognition and classification, and the bag-of-visual-words model is a widely used method for image feature representation. This paper analyzes the shortcomings of the existing models for constructing the visual dictionary, and proposes a new model for constructing the visual dictionary. First, the feature vectors are divided into the smooth class and the edge class by using the gradient variance. And then, the visual dictionaries of the two kinds of feature vectors are constructed respectively. Finally, the bag-of-visual-words model is generated according to the two kinds of visual dictionaries. Compared with the traditional method, the new method can improve the classification accuracy in the image classification experiments.

Key words: bag-of-visual-words, image classification, vector quantization, feature representation