CSIQ Video Quality Database
The CSIQ video database was developed to provide a useful dataset for the validation of objective video quality assessment algorithms. It consists of 12 high-quality reference videos and 216 distorted videos from six different types of distortion. The experiment was conducted via the SAMVIQ procedure to collect subjective ratings of quality.
If you use this database, please cite the following paper: P. V. Vu and D. M. Chandler, "ViS3: An Algorithm for Video Quality Assessment via Analysis of Spatial and Spatiotemporal Slices," Journal of Electronic Imaging, 23 (1), 01316, 2014, doi: 10.1117/1.JEI.23.1.013016.
- General information: README.txt
- Subjective ratings of quality: video_subj_ratings.txt
- Video database (download all six parts, and then use 7-zip or other ZIP software to merge/decompress the parts):
Reference and Distorted Videos
All videos in this database are in raw YUV420 format with a resolution of 832x480 pixels, a duration of 10 s, and span a range of frame rates: 24, 25, 30, 50, and 60 fps. For each reference video, we applied one of six types of distortion at three different levels of distortion. The distortion types consist of four compression-based distortion types and two transmission-based distortion types:
- H.264 compression (H.264)
- HEVC/H.265 compression (HEVC)
- Motion JPEG compression (MJPEG)
- Wavelet-based compression using the Snow codec (SNOW)
- H.264 videos subjected to simulated wireless transmission loss (WIRELESS)
- Additive white noise (AWGN)
For more information, please see [P. V. Vu and D. M. Chandler, "ViS3: An Algorithm for Video Quality Assessment via Analysis of Spatial and Spatiotemporal Slices," Journal of Electronic Imaging, 23 (1), 01316, 2014] and the README.txt file.