Research on quality of experience is strongly linked to viewers' opinion. To perform quality assessment on different aspects of multimedia content, we captured various 3D, HDR and UHD images and videos. We have decided to make the data set available to the research community free of charge. Please view the database on our website, and if you need the full resolution images and videos, please contact us. If you use these images and videos in your research, we kindly ask that you reference this website and our papers listed below.
DML-LBVS-HDR: Learning Based Visual Saliency prediction for High Dynamic Range Video
The followings are links to our DML-LBVS-HDR data sets. LBVS-HDR is a Visual Attention Model (VAM) for High Dynamic Range (HDR) video data. To access the database click here.
Disclaimer: This data set is free to be used for any non-commercial purposes.
Please cite the following papers if you are using this database for your research.
- Amin Banitalebi-Dehkordi, Y. Dong, Mahsa T. Pourazad, and Panos Nasiopoulos, "A Learning Based Visual Saliency Fusion Model For High Dynamic Range Video (LBVS-HDR)," 23rd European Signal Processing Conference, EUSIPCO 2015.
- Amin Banitalebi-Dehkordi, Maryam Azimi, Mahsa T. Pourazad, and Panos Nasiopoulos, "Saliency-Aided HDR Quality Metrics," ISO/IEC JTC1/SC29/WG11 MPEG2014/m37317, Oct. 2015, Geneva, Switzerland.
|
|
|
|
||
DML-3D-HFR: High Quality 3D Video Dataset with Different Frame Rates
The followings are links to our 3D video data sets, which were used to study the effect of frame rate on 3D video Quality of Experience. The data set contains 5 videos of the same scene with different frame rates of 24, 30, 48, and 60 fps. To access the database click here.
Disclaimer: This data set is free to be used for any non-commercial purposes.
Please cite the following papers if you are using this database for your research.
- Amin Banitalebi-Dehkordi, Mahsa T. Pourazad, and Panos Nasiopoulos, "Effect of High Frame Rates on 3D video quality of experience," 32nd International Conference on Consumer Electronics, ICCE, Jan. 2014, Las Vegas, US.
- Amin Banitalebi-Dehkordi, Mahsa T. Pourazad, and Panos Nasiopoulos, "The effect of frame rate on 3D video quality and bitrate," Springer Journal of 3D Research, vol. 6:1, pp. 5-34, March 2015, DOI 10.1007/s13319-014-0034-3.
|
|
|
|
|
|
DML-iTrack-3D dataset contains the eye-tracking data of 61 stereoscopic 3D sequences collected from 24 subjects. 27 videos are marked as the "training" videos and their corresponding fixation maps are available to download. The rest of the videos (34) are marked as the "validation" set and are used for performance evaluations. The fixation maps of the validation set is not made publicly available to conduct a fair comparison among the performance of various visual attention models. To access the database click here.
Disclaimer: This data set is free to be used for any non-commercial purposes.
Please cite the following papers if you are using this database for your research.
- Amin Banitalebi-Dehkordi, Eleni Nasiopoulos, Mahsa T. Pourazad, and Panos Nasiopoulos, "Benchmark three-dimensional eye-tracking dataset for visual saliency prediction on stereoscopic three-dimensional video," J. Electron. Imaging. 25 (1), 013008 (January 14, 2016); doi: 10.1117/1.JEI.25.1.013008.
DML-HDR: HDR Video Database
The followings are links to our High Dynamic Range (DML-HDR) video data sets, which were used in our study to evaluate various HDR image and video quality metrics. DML-HDR database includes five videos with different characteristic, which are captured by a RED SCARLET-X camera capable of capturing HDR motion footage. Each video sequence is approximately 10 seconds long with frame rate of 30 frames per second (fps) and resolution of 2048x1080. The data set contains 5 HDR videos in two different formats:
1) For each video, frames are stored in ".hdr" format. There are many software programs that can read and open this image format, for example "MATLAB" or "Picturenaut". RGBE is a lossless HDR video format, where each pixel value consists of one byte for red mantissa, one bite for green mantissa, one for the blue mantissa, and one bite for a common exponent.
2) Each video is converted to a "YUV" 12-bit file. YUV 12 bit format consists of three channels, Y for luma and U and V for Chroma. Each channel is represented by integer values between 0 and 4095 (12 bits). The process of generating the YUV files is described in our paper. "PYUV" is a typical software to read these files. Note that since conversion from the original HDR videos to YUV format results in loss of information and also color space conversion is applied in this process, we encourage using the ".hdr" file format rather than "YUV" ones.
To access the database click here.
Disclaimer: This data set is free to be used for any non-commercial purposes.
Please cite the following papers if you are using this database for your research.
- Maryam Azimi, Amin Banitalebi-Dehkordi, Yuanyuan Dong, Mahsa T. Pourazad, and Panos Nasiopoulos, "Evaluating the Performance of Existing Full-Reference Quality Metrics on High Dynamic Range (HDR) Video Content," ICMSP 2014: XII International Conference on Multimedia Signal Processing, Nov. 2014, Venice, Italy.
- Amin Banitalebi-Dehkordi, Maryam Azimi, Mahsa T. Pourazad, and Panos Nasiopoulos, "High Dynamic Range Video Compression Using HEVC and H.264/AVC Standards," 10th International Conference on Heterogeneous Networking for Quality, Reliability, Security and Robustness, QSHINE, Aug. 2014, Greece.
|
|
|
|
|
|
|
||
DML-iTrack-HDR dataset is an eye tracking dataset for HDR video. Ten HDR video sequences were shown on a Dolby prototype HDR display, and the eye movements of the viewers were recorded by a SensoMotoric Instruments (SMI) eye tracker. These videos were watched by 18 individuals, and fixation density maps (FDM) were obtained for each frame.To access the database click here.
Disclaimer: This data set is free to be used for any non-commercial purposes.
Please cite the following papers if you are using this database for your research.
- E. Nasiopoulos, Y. Dong, and A. Kingstone, "Evaluation of High Dynamic Range Content Viewing Experience Using Eye-Tracking Data," 10th International Conference on Heterogeneous Networking for Quality, Reliability, Security and Robustness, QSHINE, Aug. 2014, Greece.
- Yuanyuan Dong, Elini Nasiopoulos, Mahsa T.Pourazad, Panos Nasiopoulos, "High Dynamic Range Video Eye Tracking Dataset," 2nd International Conference on Electronics, Signal processing and Communications, ESPCO, Nov 2014, Greece.
- Amin Banitalebi-Dehkordi, Yuanyuan Dong, Mahsa T. Pourazad, and Panos Nasiopoulos, "A Learning Based Visual Saliency Fusion Model For High Dynamic Range Video (LBVS-HDR)," 23rd European Signal Processing Conference, EUSIPCO 2015.
Clip Snapshot | FDM |
|
|
|
|
The motivation of creating this database was to quantify the effect of a few important parameters to 3D quality. These parameters include d_min (the distance between the cameras and the closest point captured in the stereoscopic image pair), d_max (the distance between the camera and the background), and d_obj (the distance between the camera and the main object of interest). Note the interocular distance between the two parallaled cameras were 77mm for all images in database 1.
Disclaimer: This data set is free to be used for any non-commercial purposes.
Please cite the following papers if you are using this database for your research.
- Lino E. Coria, Di Xu, Panos Nasiopoulos, "Quality of Experience of Stereoscopic Content on Displays of Different Sizes: A Comprehensive Subjective Evaluation," IEEE International Conference on Consumer Electronics ICCE 2011, Las Vegas, NV, USA, January 9-12, 2011, pages 778-779.
- Di Xu, Lino Coria, Panos Nasiopoulos, "Guidelines for Capturing High Quality Stereoscopic Content Based on a Systematic Subjective Evaluation," IEEE International Conference on Electronics, Circuits, and Systems, ICECS 2010, Athens, Greece, December 12-15, 2010, pages 166-169.
left view and right view |
left view and right view |
left view and right view |
left view and right view |