4D studio for future internet: Improving foreground-background segmentation

The 4D Reconstruction Studio developed at MTA SZTAKI enables real-time creation of dynamic 3D models for moving articulated objects such as humans. Image segmentation is a critical step of the 4D reconstruction pipeline as it strongly influences the quality of the models in terms of both geometry and texturing. This paper presents the efficient image segmentation approach we developed for the studio. We discuss our novel solutions that improve the precision and robustness of segmentation against illumination variations and shadows.