Shadow Removal from a Real Image Based on Color Clustering

DOI Web Site 1 Citations 11 References
  • Baba Masashi
    Faculty of Information Sciences Systems, Hiroshima City University
  • Mukunoki Masayuki
    Faculty of Information Sciences Systems, Hiroshima City University
  • Asada Naoki
    Faculty of Information Sciences Systems, Hiroshima City University

Bibliographic Information

Other Title
  • 色分布のクラスタ解析による実写画像からの陰影除去
  • イロ ブンプ ノ クラスタ カイセキ ニ ヨル ジッシャ ガゾウ カラ ノ インエイ ジョキョ

Search this article

Description

This paper proposes a method to remove shadows from a real image based on RGB color space analysis. The algorithm consists of five steps. First, all pixel colors are mapped on the RGB space, and the color distribution is segmented by using clustering method. Second, a pair of clusters that are almost aligned on a color vector is found. We consider the darker cluster as shadow area and the lighter as sunshine one. Next, we perform the color adjustment between them by modifying the direction of shadow cluster to be that of sunshine one. Fourth, the color average and variance of the shadow cluster are adjusted to be the same as those of the sunshine one. Finally, we apply a smoothing filter to correct the boundary discontinuity that sometimes occurs around fuzzy boundary between shadow and sunshine areas. Experimental results show that our algorithm worked successfully inreal situations.

Journal

Citations (1)*help

See more

References(11)*help

See more

Keywords

Details 詳細情報について

Report a problem

Back to top