3D Reconstruction Using Light Field Camera

• Main Author: Wee, Kelvin Ting
• Format: Final Year Project
• Language: English
• Published: IRC 2015
• Subjects: TJ Mechanical engineering and machinery
• Online Access: http://utpedia.utp.edu.my/15682/1/DISSERTATION_KELVIN%20TING%20WEE_ME_14954.pdf; http://utpedia.utp.edu.my/15682/
• Institution: Universiti Teknologi Petronas

3D reconstruction is important as a method to represent the real world environment using a 3D digital model. The emergence of the Lytro light field camera in the market has opened up new possibilities for researchers to explore 3D reconstruction with this easily obtained off-the-shelf product. By using depth information contained in the camera, 3D reconstruction of real life objects is made possible. However, despite its huge potential, 3D reconstruction based on light field technology is still insufficiently explored. In this work, a map is obtained by using two different responses of the image, namely defocus and correspondence response, and combining both responses to get a clearer and better depth map. In the beginning stage of research, one image with a fixed point of focus is being selected as object of study and exported in multiple file formats. Some of it contains all the light field information of Lytro image, while others contain selective information such as depth data or representation of 2D image. At the initial stage, a preliminary depth map was obtained but the depth representation was not clear and obvious. In the end a 3D depth map that has the outline and shape of a real object studied was generated. It was later found out that defocus analysis can be improved by reducing the defocus analysis radius. All in all, a 3D depth map can be successfully obtained from light field picture through computations in MATLAB code.