基于多维深度线索的视频2D转3D系统设计

his paper presents a multi-dimensional depth cues based video 2D to 3D system design. The program leads through the integration movement, perspective cues and clues to get the final texture depth information, the use of virtual viewpoint image synthesis algorithm synthesis and empty eye fix and then render 3D effects. Among them, the campaign trail in accordance with the movement of the object is divided into two categories namely objects in the scene background and foreground, dark-depth information that is non-Ming; perspective clues to find the images off-line or off-point, thereby generating a smooth gradient of varying depth information; texture leads to a color image into a grayscale image, generating a luminance information image; different weights to generate the final depth information to the depth information in three different dimensions, and depth information smoothing filter. Multidimensional depth information depth cues fusion method is more consistent with the generated stereoscopic perception system of the human eye, its low computational complexity, while the algorithm for different types of 2D scene information can achieve better depth information. Experimental results show that compared with other systems consider only a single dimension of depth information compared to the multi-dimensional depth cues based video 2D to 3D system presents a wide range of 3D effects in terms of adaptability and smooth playback has better performance, better real-time .