Gradientベースの特徴抽出 -SIFTとHOG-  [in Japanese] Gradient-Based Feature Extraction -SIFT and HOG-  [in Japanese]

Access this Article

Search this Article

Author(s)

Abstract

Scale-Invariant Feature Transform(SIFT) は,特徴点の検出と特徴量の記述を行うアルゴリズムである.検出した特徴点に対して,画像の回転・スケール変化・照明変化等に頑健な特徴量を記述するため,イメージモザイク等の画像のマッチングや物体認識・検出に用いられている.本稿では, SIFT のアルゴリズムについて概説し,具体例として SIFT を用いたアプリケーションや応用手法への展開について紹介する.また,SIFT と同様に gradient ベースの特徴抽出法であるHistograms of Oriented Gradients(HOG)のアルゴリズムとその応用例として人検出についても紹介する.Scale-Invariant Feature Transform(SIFT) is an approach for detecting and extracting local feature descriptors that are reasonably invariant to changes in illumination, image noise, rotation, scaling, and small changes in viewpoint. Because the SIFT algorithm can describe characteristics of feature points that are invariant to scale and rotation changes, it has been used for image matching such as image mosaicing and generic object recognition. In this paper, we describe the SIFT algorithm and introduce applications that use it. We also describe another algorithm called "Histograms of Oriented Gradients(HOG)"which is based on gradient feature extraction similar to the SIFT algorithm. We also introduce an example of how HOG can be used for people detection.

Scale-Invariant Feature Transform(SIFT) is an approach for detecting and extracting local feature descriptors that are reasonably invariant to changes in illumination, image noise, rotation, scaling, and small changes in viewpoint. Because the SIFT algorithm can describe characteristics of feature points that are invariant to scale and rotation changes, it has been used for image matching such as image mosaicing and generic object recognition. In this paper, we describe the SIFT algorithm and introduce applications that use it. We also describe another algorithm called "Histograms of Oriented Gradients(HOG)" which is based on gradient feature extraction similar to the SIFT algorithm. We also introduce an example of how HOG can be used for people detection.

Journal

  • CVIM

    CVIM 2007(87(2007-CVIM-160)), 211-224, 2007-09-04

    Information Processing Society of Japan (IPSJ)

References:  35

Cited by:  65

Codes

  • NII Article ID (NAID)
    110006403842
  • NII NACSIS-CAT ID (NCID)
    AA11131797
  • Text Lang
    JPN
  • Article Type
    Journal Article
  • ISSN
    09196072
  • NDL Article ID
    8938070
  • NDL Source Classification
    ZM13(科学技術--科学技術一般--データ処理・計算機)
  • NDL Call No.
    Z14-1121
  • Data Source
    CJP  CJPref  NDL  NII-ELS  IPSJ 
Page Top