VQEG Column: Finalization of Recommendation Series P.1204, a Multi-Model Video Quality Evaluation Standard — The New Standards P.1204.1 and P.1204.2

This column introduces the now completed ITU-T P.1204 video quality model standards for assessing sequences up to UHD/4K resolution. Initially developed over two years by ITU-T Study Group 12 (Question Q14/12) and VQEG, the work used a large dataset of 26 subjective tests (13 for training, 13 for validation), each involving at least 24 participants rating sequences on the 5-point ACR scale. The tests covered diverse encoding settings, bitrates, resolutions, and framerates for H.264/AVC, H.265/HEVC, and VP9 codecs. The resulting 5,000-sequence dataset forms the largest lab-based source for model development to date. Initially standardized were P.1204.3, a no-reference bitstream-based model with full bitstream access, P.1204.4, a pixel-based, reduced-/full-reference model, and P.1204.5, a no-reference hybrid model. The current record focuses on the latest additions to the series, namely P.1204.1, a parametric, metadata-based model using only information about which codec was used, plus bitrate, framerate and resolution, and P.1204.2, which in addition uses frame-size and frame-type information to include video-content aspects into the predictions.