Skip to Main Content
Depth map flickering of object in the scene makes human feel fatigue for 2D-to-3D video conversion on 3DTV. In this paper, we present a fast and robust scheme to generate depth map from monocular video without depth map flickering. The proposed method is to divide input video sequence into the cuts and assign initial depth map and segment object in the image by using color and motion history. Mixed information of color and motion is applied to diminish depth error. The experimental results show that our scheme achieves real-time performance and reduce human eye fatigue on 3DTV.