计算机工程 ›› 2013, Vol. 39 ›› Issue (5): 212-217.doi: 10.3969/j.issn.1000-3428.2013.05.047

• 人工智能及识别技术 • 上一篇    下一篇

方向梯度直方图及其扩展

傅红普1,2,邹北骥1   

  1. (1. 中南大学信息科学与工程学院,长沙 410086;2. 湖南第一师范学院信息科学与工程系,长沙 410205)
  • 收稿日期:2012-05-11 出版日期:2013-05-15 发布日期:2013-05-14
  • 作者简介:傅红普(1973-),男,副教授、博士研究生,主研方向:机器视觉;邹北骥,教授、博士
  • 基金项目:
    湖南省教育厅科学研究基金资助项目(12C0591);湖南第一师范学院科研基金资助项目(XYS07N12)

Histogram of Oriented Gradient and Its Extension

FU Hong-pu 1,2, ZOU Bei-ji 1   

  1. (1. School of Information Science and Engineering, Central South University, Changsha 410086, China; 2. Department of Information Science and Engineering, Hunan First Normal University, Changsha 410205, China)
  • Received:2012-05-11 Online:2013-05-15 Published:2013-05-14

摘要: 介绍方向梯度直方图(HOG)及其扩展。结合HOG的计算步骤阐述其内在理念和捕获的主要图像特征,指出其局限性,论述对HOG的重要扩展,包括多尺寸HOG、梯度方向直方图金字塔、分析降维HOG,给出对各扩展使用积分图加速计算的方法以及其他改善方法,从措施、复杂度、维数、思想等方面对各HOG改善方法进行比较,给出HOG未来的研究方向。

关键词: 局部特征描述符, 方向梯度直方图, 积分图, 梯度方向直方图金字塔, 降维分析

Abstract: This paper surveys Histogram of Oriented Gradient(HOG) and its extension. Combining with the computing step of HOG, this paper discusses its underline idea, the main image features captured by it and its limitations. It discusses several important extensions of HOG, including variable-size HOG, Pyramid of Histogram of Oriented Gradient(PHOG) and analytic dimension reduction HOG. And it also describes the way to speedup the computing of these extensions with integral image. It gives other possible ways to improve HOG and the possible directions to improve the ability of HOG, and makes the comparison of HOG improving methods with measure, complexity, dimension and idea. Several valuable research directions are proposed.

Key words: local feature descriptor, Histogram of Oriented Gradient(HOG), integral imag, Pyramid of Histogram of Oriented Gradient(PHOG), dimension reduction analysis

中图分类号: