Stereoscopic image generation from 2D image based on object feature analysis

In order to obtain a better stereoscopic image from a 2D image, it needs to generate an accurate depth-map from the 2D image. This paper proposes a method which generates stereoscopic images for outdoor scene by based on object feature analysis. First, it detects vanishing lines and vanishing point on an image. Second, it segments the objects on the image by Hill-climbing based segmentation and makes initial depth-map. Finally, it accurates the initial depth-map for the image based on object feature analysis. The final depth-map is used for generation of the stereoscopic image. Experimental results of the method show that the synthetic stereo image pairs are close to the original stereo image pairs.