ヒストグラム拡張とクラスタリングによる視界不良画像の鮮明化に関する研究

カテゴリ: 論文誌(論文単位)

グループ名: 【C】電子・情報・システム部門

発行日: 2016/10/01

タイトル(英語): Contrast Enhancement Based on Histogram Stretch and Clustering of Brightness in Invisible Scene

著者名: 瓦吹　大（弘前大学大学院理工学研究科），小野口　一則（弘前大学大学院理工学研究科）

著者名(英語): Hiroshi Kawarabuki (Graduate School of Science and Technology, Hirosaki University), Kazunori Onoguchi (Graduate School of Science and Technology, Hirosaki University)

キーワード: 画像処理，コントラスト強調，ヒストグラム拡張，クラスタリング　　image processing，contrast enhancement，histogram stretch，clustering

要約(英語): In this paper, we propose a novel contrast enhancement algorithm. The contrast enhancement is an important for preprocessing of image processing because video processing for outdoor surveillance is always influenced by the weather. Most of conventional algorithms premise that images are clear. We try to highlight the invisible objects obscured by weather, such as rainy, snowfalls, night vision, and so on.There are two critical problems in field of contrast enhancement. One is to preserve brightness and the other is to suppress false color. We employ the contrast enhancement algorithm based on histogram stretch because histogram stretch is able to preserve brightness of input images. In order to greatly improve contrast of images, we combine the clustering with histogram stretch. In addition, we divide color into brightness and saturation because false color is caused by an over-enhancement of saturation. So our algorithm preserves brightness and suppresses the false color generation. Furthermore, in various weathers, our algorithm can obtain the stable images. To this end, our algorithm estimates the degree of histogram stretch in any images automatically. Experimental results show that the proposed algorithm can make images clear effectively, even if original images are unclear by any weather.

本誌: 電気学会論文誌C（電子・情報・システム部門誌） Vol.136 No.10 （2016） 特集：電気関係学会四国支部連合大会

本誌掲載ページ: 1473-1482 p

原稿種別: 論文／日本語

電子版へのリンク: https://www.jstage.jst.go.jp/article/ieejeiss/136/10/136_1473/_article/-char/ja/