Akademik Veri Yönetim Sistemi
23 - 26 Aralık 2021, ss.17
Forming efficient 3 Dimensional (3D) video quality prediction models can be considered as one of the most important research areas to enable enhanced multimedia services. However, there exists no standardized model in the literature to provide efficient prediction of the 3D video quality. Therefore, the quality assessment models used for 2 Dimensional (2D) videos are currently exploited for the 3D video quality prediction as well. Nevertheless, using the 2D video quality assessment models can only be in compliance with the 3D video quality prediction if and only if Human Visual System (HVS) related characteristics are taken into consideration. Considering this fact, Peak Signal to Noise Ratio (PSNR) is improved to form a 3D video quality prediction model relying on the HVS related characteristics. While forming this model, first of all, an abstraction filter is used to emphasize HVS related characteristics in different 3D videos. Then, PSNR is utilized to measure the qualities of the abstracted 3D videos. Following this, contrast and motion information values of the 3D videos are measured using devised models. Moreover, subjective tests are carried out to assess the depth perception of different users towards the 3D videos used in this study. At the last step of the model development, the measured contrast and motion information values are integrated with the PSNR values with the assistance of the subjective test results to devise the proposed model. In order to assess the performance of the proposed model, correlation coefficient values are computed with the original PSNR values. The results reveal that the proposed model is quite efficient to measure the depth perception of the 3D videos in comparison to the original PSNR.