| With the development of science technology and the improvement of people’s living qualities, 3D video is applied in our daily life, especially being popular in home entertainment. Nowadays, 3D video is not enough to meet the requirements of future 3D video market. So it has become a focus that 2D videos is transformed to 3D videos. At present, there are two key techonologies for Converting 2D to 3D video. O ne is to generate depth map of 2D images, and the other one is to convert the depth map to 3D video which can be displayed on the 3D display device.It is a difficult work for the generation of depth map. Due to the scene is always composited by moving foreground and static background, as well as the visual characteristics of human, a generation method of depth map based on motion information and geometric perspective is adopted in this thesis. For the moving foreground, the optical flow method is used to obtain the motion vector, and then the motion vector is converted to depth map of foreground according to the relationship between motion vector and depth. For the scnes of static background, firstly, sobel edge detection and Hough transform are used to extract the off- line and vanishing point of the scen, and then the depth assignments along the off- line direction among background are excuted according to the position of the vanishing point. Thus the depth map of background is generated. Finally, the foreground and background depth map are fusioned to get the final depth map. Experimental results show that the method based on motion information and geometric perspective can get high-quality depth map.For converting the depth map to 3D video, a 3D conversion method based on disparity map is adopted in this thesis, which the original 2D video and depth map are used to genertate 3D vedio. For the hole problem, a hole fill method based on the disparity map is proposed to fill the hole effectively, and an effective method of transform 2D video into 3D video is obtained. |
PDF Full Text Request