Foreground segmentation with sudden illumination changes using a shading model and a Gaussianity test
Ka Ki Ng, Satyam Srivastava, Edward J. Delp · International Symposium on Image and Signal Processing and Analysis · 2011
In this paper, we propose a simple method for foreground segmentation based on a “Gaussianity test” and a shading model. The proposed method works under a hierarchical framework that combines a block based and a pixel based processing. The first step is a block-level classification based on the intensity differences and intensity ratios of a background model and the current frame. We then use pixel-wise adaptive background subtraction on the foreground classified blocks to obtain the foreground mask. We test our proposed method on several sequences with indoor scenes with extreme sudden illumination changes, and outdoor scenes under strong sunlight, waving tree leaves, and walking pedestrians. The method is shown to be robust to extreme sudden illumination changes and the presence of relatively small scene clutter motion (e.g. waving tree branches and leaves).