| In recent years,3D display devices and 3D video have been rapidly spread with the rapid development of 3D display technology.3D video has been widely welcomed, because its stereoscopic effects can improve people’s visual experience. However, the shortage of 3D video contents restricts the further development of 3D display technology.2D to 3D conversion technology has been deeply researched, because it can convert existing 2D video contents into 3D video contents. In 2D to 3D, Estimating corresponding depth map based on the known 2D view is one of the key steps. In recent years, depth propagation technique has been widely used and deeply studied as an important technology in semi-automatic 2D-3D conversion. It is mainly based on the correlation between video frames. First, Specific frames are selected as keyframes in video sequences. Depth values of key-frame are manually assigned or known in advance. Then the depth map of non-key frames can be obtained according to the correlation between frames. The application of depth propagation makes the quality of depth map better in 2D to 3D with fewer human efforts. The existing depth propagation algorithm can be generally divided into four categories:method based on motion estimation and bilateral filter, method based on optimization modeling, method based on image segmentation and method based on machine learning. The existing depth propagation algorithm has achieved good results, but there are still some problems, such as blocking artifacts and blur, to be improved.Thanks to development of 3D display technology, multi-view display technology evolves and improves the user’s visual experience. However, the increased amount of data brought by multi-view video makes the compression, storage and transmission of 3D video content face a challenge. In a traditional multi-view video coding system, which is called 3DV systems, the color maps and depth maps of multiple views need to be coded. The coding efficiency and bandwidth should be limited. Among the multi-view video coding standards studied in recent years, a variety of video coding techniques have been applied and expanded.The main work of this paper is as follows:(1) A depth propagation algorithm based on adaptive motion estimation is proposed because of the shortcuts of existing algorithms. The main idea of this algorithm is that different areas of the scene take different motion estimation methods. For background area, the depth is so smooth that bi-directional overlapped block motion compensation method is applied. For the foreground area with rich depth information and background area with inaccurate motion vector, the method based optical flow field is used. The algorithm can be applied to the case of complex scenes, and the generated depth map has better quality.(2) An inter-view depth map generation method based on inter-view depth propagation technique is proposed for the increase of transmission data brought by multi-view video display techniques. It can be applied to depth coding of multi-view video coding in 3DV system to save transmission bandwidth. The algorithm includes three parts:inter-view depth propagation,3D image warping and depth value selection. According to the method, except the color maps of all views, only depth maps of few key views and binary matrixes corresponding to non-key views should be transmitted, and all depth maps can be acquired in decoder. Experimental results show that the quality of depth map obtained by this method is similar with depth map obtained by normal video reconstruction. Meanwhile, this method can significantly reduce the transmission bandwidth of depth coding. |
PDF Full Text Request