Objective quality measures comparison of impaired 3D video sequences from the UC3D database

In this study we analyze the correlation between different image and video quality measures and corresponding subjective scores computed on 20 3D texture+depth video sequences from the University of Coimbra 3D (UC3D) video database. Quality metrics were tested on (degraded) depth information and synthesized views. The results show that in the case of measures based on depth information, the best correlation is achieved by SSIM. For the case of measures computed on the synthesized view, the highest correlation is obtained when using PSNR. Overall results show that quality measures computed on the synthesized view correlate better with subjective (DMOS) scores than measures based on the depth information.