Depth Map Generation Based On Object Geometric Models

In recent years, 3D video film and television, augmented reality and virtual reality have been developed more and more rapidly. Stereo video production, as its core technology, is getting more and more attention. 2D-3D video conversion can convert 2D image into 3D image through specific algorithm without extra filming equipment. It can effectively transfer existing 2D video resources into 3D video contents to meet the large demand for stereo video. However, due to the complexity of image content, limits like low automation, need a lot of manual participation and high production costs are still remaining in current algorithms. Therefore, research on 2D-3D video conversion technology, improve the quality of 3D video imaging have important value.Based on the above considerations, we noticed the importance of the depth information of foreground object to stereo visual effects. Focus on the calculation and optimization method of image foreground objects depth information. In addition, we also put forward a new algorithm of background depth calculation. The main work of this paper is as follows:1. This article proposed a new foreground depth assignment method for the situation that the image foreground object has some geosemantic information. Our algorithm decompose foreground object into basic geometric model and the depth information of foreground objects is obtained by these basic models. We also established some common basic geometries' 3D model and implemented depth information calculation and optimization method based on image segmentation and basic geometric models.2. We designed and implemented an interactive depth map generation system. As there are still many shortcomings in automatic depth map generation methods, adding a certain amount of human interaction can amend the defective part timely. The system achieves the functions of separation of foreground and background, foreground background depth assignment through simple artificial interaction, and is used to generate high quality depth map.