STEREOSCOPIC CONVERSION OF MONOSCOPIC VIDEO USING EDGE DIRECTION HISTOGRAM

摘要

In this paper, we propose an algorithm for converting a monoscopic 2D image into stereoscopic 3D image by generating depth information based on vanishing points. The distance between two parallel lines in a 3D space becomes narrower as they are further away from the perspective of the viewer, until they finally meet at one point called a vanishing point. The viewer is able to estimate this vanishing point with geometrical features in an image, and perceives depth with the relative position of the vanishing point. In the proposed algorithm, depth information in each image is generated by the relative position of a vanishing point estimated by an edge directional histogram and one of the predefined initial depth maps is selected with this depth information. A final depth map is then generated by properly combining the selected initial depth map and vertical linearity depth map obtained with an assumption that the lower parts of an image usually have more depth. The left and right images for the stereoscopic 3D image are then generated with the final depth map. Experimental results show that the proposed algorithm produces a good quality of stereoscopic images with less computational power.