¡¡Chinese Journal of Computers   Full Text
  TitleAn Image Segment Method Based on Color Invariance of Physical Reflection Model
  AuthorsCAI Xun MENG Xiang-Xu HAO Xing-Wei LIAN Li JIANG Long
  Address(Human-Interaction & Virtual-Reality Research Department, School of Computer Science and Technology, Shandong University, Jinan 250061)
  Year2009
  IssueNo.2(282¡ª287)
  Abstract &
  Background 		Abstract This paper proposes a new background segmentation method based on Phong reflection model for video surveillance system. The main goal of this method is to reduce the affections of the non-uniform illumination and specular reflection. The steps of this method include: First, the colour invariance of diffuse reflection was deduced and the diffuse component of the model was estimated; second, the specular factors and specular exponent of the specular component was calculated by following differential rules; third, based on the established model, the background segment is accomplished. The experimental results and the evaluation show that this method can not only be robust to the non-uniform lighting and the specular reflection of background surface, but also be applicable for real-time video surveillance system.
Keywords color invariance; threshold segmentation; computer vision, video surveillance; Phong illumination model
Background The study in this paper is based on my researches of Ph.D. dissertation and the project of Shandong Natural & Science Foundation. In video surveillance system, color is one of the most important features for object recognition. However, color is often affected by illumination change, so how to reduce the illumination change has become a very important problem in computer vision. During the study, the authors find Phong illumination model which oftern be used in computer graphic is a good physical model to simulate the illumination of single light source. So they begin to study a new method to reduce the non-uniform illumination and specular effects. Compared with most methods which use HSV space or other chromatic spaces to reduce the effects of small illumination change, our method which combined the physical illumination model with the statistics method can not only be robust to the non-uniform lighting and the specular reflection of background surface, but also be applicable for real-time video surveillance system.