2D to 3D Video Conversion Based on Interframe Pixel Matching

This paper presents a new method for 2D to 3D video conversion based on motion information. This method exploits the different relations of pixels matching after motion estimation to generate depth map of the 2D video. On the edge of moving object, pixels matching degree is utilized to judge that the pixel belongs to foreground or background. The contour of moving object will become more distinct. Furthermore, to the pixel which hasn't get correct matching, corresponding approach is given according to the object motion. The experimental results show that the sawtooth effect of depth map significantly reduced by proposed method. The quality of 3D video is improved.