Main vertical objects detecting algorithm for real-time 2D-to-3D conversion

Lacking of enough 3D contents has become the bottleneck of 3D technology development. 2D-to-3D conversion can not only solve this problem, but also be compatible with the 2D video format. However, most of the real-time 2D-to-3D conversion methods are limited to some special images/videos. This article proposes a scheme that can be applied to all kinds of images/videos. Firstly, the motion vectors are obtained by using bi-directional motion estimation. Combining the motion information and the RGB values of the image, a new image segmentation technology is proposed. With each segmented region, the main vertical objects are detected. Finally, the main vertical objects are assigned the same disparity value based on its bottom vertical position. The disparities of other pixels are determined on its vertical position, which can be seen as a safety methods that uses 3D cues implicitly. The experimental results show its stronger 3D effect.