An Automatic Depth Map Estimation System for Single-view Outdoor Images

• Main Authors: Chih-Wei Chen, 陳志瑋
• Other Authors: Jar-Ferr Yang
• Format: Others
• Language: en_US
• Published: 2009
• Online Access: http://ndltd.ncl.edu.tw/handle/81675939549874263964

碩士 === 國立成功大學 === 電腦與通信工程研究所 === 97 === With 3D display technology improvement, people now can perceive the 3D effect by 3D display without wearing 3D glasses. We have a huge amount of 2D video contents; however, due to 3D contents are not sufficient enough, the 2D to 3D conversion technique becomes more and more important to promote 3D displays in the near future. In the literatures, several researches proposed depth map estimation methods if multi-view images are available. However, for those thousands 2D videos, the most demanding method is to estimate the depth map from original 2D images. In this thesis, we propose an automatic depth map generation method from general single-view outdoor background images. In the proposed automatic depth map generation system, we first develop some image landscape extractions which can detect some outdoor landscapes, such as sky, ground line, and tree, and then segment the desired objects. After segmentation, we assign the depth value to the extracted objects based on their spatial characteristics. Experimental results show that the proposed method can effectively extracted the desired landscape objects and generates their reasonable depth maps for outdoor images.