ITU-T P.1204 is a family of standardized video-quality models published by the International Telecommunication Union in 2020 for adaptive HTTP streaming over reliable transport at resolutions up to 4K. It is not one model but three, distinguished by what they are allowed to look at: P.1204.3 is bitstream-based and no-reference, parsing the compressed stream's quantization parameters, motion vectors, and coefficient statistics with no original needed; P.1204.4 is pixel-based and uses the reference (a reduced-reference design behaving close to full-reference); and P.1204.5 is a hybrid no-reference model combining bitstream data with received pixels. All output a Mean Opinion Score on a 1-to-5 scale and were trained in the P.NATS Phase 2 competition with ITU-T Study Group 12 and VQEG. The headline result is that bitstream P.1204.3 matched human opinion above full-reference VMAF in its own validation (Pearson 0.942 versus 0.873). The catch: P.1204.3 is defined only for H.264, HEVC, and VP9, and feeds the P.1203 session-QoE framework.
