Ten uncompressed, high definition (1920×1088) videos are used as source videos. One video which contains animations were rendered directly from 3D models, while other videos which contains natural scenes were shot with professional, high end equipment and converted to digital format with utmost care. The videos shot by us were recorded in a progressive RAW format at 25 fps without any compression. Distorted sequences are generated using 4 types of distortions: MPEG 2 compression, Dirac wavelet compression, H.264 compression and packet loss on the H.264 streaming through IP networks. MPEG 2 is a traditional standard in broadcasting society, H.264 is rapidly gaining popularity due to its superior compression efficiency, and Dirac wavelet is also competitive codec employed by BBC (British Broadcast Corporation). To ensure an enough span of video qualities, the distortion strengths were diversified. 42 paid observers participated in the subjective test, including 11 females and 31 males. Each observer assessed 10 source videos and 128 distorted videos. The single-stimulus quality scale test method (ACR) was used, where each video (including the reference) occurred once in a random order yet the two successive videos must come from different source videos so as to remove contextual and memory effects in quality evaluation.
Data is accessible via FTP (Username: videodb, Password: sequence2011) Link: ftp://ivp.ee.cuhk.edu.hk/ Detailed description of the experiment is accessible via HTTP: Link: http://ivp.ee.cuhk.edu.hk/research/database/subjective/Detailed%20Information%20about%20the%20IVP%20Subjective%20Quality%20Video%20Database.pdf