MiX-LFQDB: MIUN-Xidian Light Field Quality Database for Compressed Light Field Images using Learning-based vs. Conventional Methods doi-10-5281-zenodo-16778671-0 https://doi.org/10.5281/zenodo.16778671 Swedish National Data Service Svensk nationell datatjänst Landing page MiX-LFQDB: MIUN-Xidian Light Field Quality Database for Compressed Light Field Images using Learning-based vs. Conventional Methods doi-10-5281-zenodo-16778671-0 https://doi.org/10.5281/zenodo.16778671 Zerman, Emin Takhtardeshir, Soheib TRIOUX, Anthony Olsson, Roger Sjöström, Mårten Swedish National Data Service Svensk nationell datatjänst Landing page This database was created by a joint effort from the Realistic 3D research group at Mid Sweden University (Sundsvall, Sweden) and the School of Telecommunications Engineering in Xidian University (Xi'an, China). The database details are explained thoroughly in the publication, which is accepted (and to appear in the proceedings of) the 27th IEEE International Workshop on Multimedia Signal Processing (MMSP) in 2025.  You can use this database in your work under the Creative Commons Attribution 4.0 International (CC-BY 4.0) licence, provided that you cite the database as below: Zerman, E., Takhtardeshir, S., Trioux, A., Qin, J., Wu, W., Olsson, R., & Sjöström, M. (2025). Subjective Visual Quality Assessment of Compressed Light Field Images: Learning-based vs. Conventional Methods. The 27th IEEE International Workshop on Multimedia Signal Processing (MMSP).DOI: (To be updated after publication) BibTeX: @inproceedings{zerman2025subjective  title        = {Subjective Visual Quality Assessment of Compressed Light Field Images: Learning-based vs. Conventional Methods},  author       = {Zerman, Emin and Takhtardeshir, Soheib and Trioux, Anthony and Qin, Jianlong and Wu, Wenjie and Olsson, Roger and Sj{\"o}str{\"o}m, M{\aa}rten},  booktitle    = {The 27th IEEE International Workshop on Multimedia Signal Processing (MMSP)},  year         = {2025},  organization = {IEEE}} This database contains 85 light field stimuli, rendered as pseudo-video sequences with a spiral trajectory, and the subjective quality scores collected by 40 people in two different countries (19 in Mid Sweden University, Sweden; and 21 in Xidian University, China). The 85 LF stimuli were generated from 5 source LFs using 4 different LF compression methods, comprising two conventional methods (H.265/HEVC and JPEG Pleno) and two learning-based methods (RLVC and EF-VAE). Access to data through an external actor. Data are freely accessible. Åtkomst till data via extern aktör. Data är fritt tillgängliga. openAccess