作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2009, Vol. 35 ›› Issue (14): 212-214. doi: 10.3969/j.issn.1000-3428.2009.14.074

• 人工智能及识别技术 • 上一篇    下一篇

改进的Knockout自然景物提取算法

窦 燕,肖 凤,孔令富,冯丙春   

  1. (燕山大学信息科学与工程学院,秦皇岛 066004)
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2009-07-20 发布日期:2009-07-20

Improved Knockout Natural Image Matting Algorithm

DOU Yan, XIAO Feng, KONG Ling-fu, FENG Bing-chun   

  1. (College of Information Science and Engineering, Yanshan University, Qinhuangdao 066004)
  • Received:1900-01-01 Revised:1900-01-01 Online:2009-07-20 Published:2009-07-20

摘要: 针对Knockout算法中颜色估计模型只适用于光滑图像的问题,提出一种新的颜色估计模型。采用区域生长法将未知区域划分为若干个子区域,以距离每一个子区域最近的前景和背景轮廓线上的点为中心做矩形块,分别计算2个矩形块内前景和背景像素点的颜色加权平均值,将其作为该子区域的前景和背景颜色分量。实验结果表明,改进后的颜色估计模型在图像边缘附近颜色相差较大的自然景物提取中有较好表现。

关键词: 蓝屏景物提取, 自然景物提取, 透明度( 值), 前景, 背景

Abstract: Aiming at the problem which is appeared in Knockout algorithm that the color estimate model is only applied to smooth image, a new color estimate model is proposed. The unknown region is divided into many subregions by region growth algorithm, rectangular block is made out based on the point which is nearest to the subregion in the foreground and background contour line. The weighted average values of foreground and background pixels color in rectangular block are computed respectively. The values are taken as the foreground and background color components. The improved color estimate model has better behavior in the natural image matting for the image which the edge color is very different.

Key words: blue screen image matting, natural image matting, transparency(α value), foreground, background

中图分类号: