Foreground segmentation under sudden illumination changes by feature fusion

To address the challenging problem of robust background subtraction under sudden illumination changes, a novel foreground segmentation method based on feature fusion was proposed. The method consisted of three stages. First, a scene model through integrating the global illumination function into the framework of Gaussian mixture models was built. Second, three kinds of illumination invariant features, i. e. zero mean normalized cross -correlation (ZNCC), textures, and contours, were extracted from the current frame image. Third, the illumination invariant features were combined for foreground segmentation in two steps. Specifically, the ZNCC and textures were combined in the first step, and the contour was integrated in the second step. The experimental results show that the proposed method can effectively improve the accuracy and robustness of foreground segmentation.