Foreground Segmentation by Integrating Color Modeling and Difference

• Main Authors: Shih-Ming Lin, 林世民
• Other Authors: Shih-Shinh Huang
• Format: Others
• Language: zh-TW
• Published: 2009
• Online Access: http://ndltd.ncl.edu.tw/handle/70542147306158962832

碩士 === 國立高雄第一科技大學 === 電腦與通訊工程所 === 97 === This paper provides a foreground detection algorithm that combines color model with color difference. The color model firstly transforms image color space to a new color space CIELCH which is able to manifest the characteristics of color. The color distribution of the background scene is then represented by mixture of Gaussians. In order to avoid the stationary foreground objects being classified to background mistakenly, this research also builds up the color difference model based on color difference variation between background image and input image. This promotes foreground detection accurately that through color model foreground contours locally to combine the foreground regions of color difference. Finally, several video sequences both in indoor and outdoor environments are used to validate our proposed detection algorithm whose results are more accurate and stable than traditional one.