Perceptual Quantization

Description

Perceptual Quantization (PQ), formally standardized as SMPTE ST 2084 and adopted by 3GPP for media delivery, is an Electro-Optical Transfer Function (EOTF). An EOTF defines the mathematical relationship between the digital code values in a video signal and the absolute luminance levels (in nits or candelas per square meter) produced by a display. Unlike traditional gamma curves used for Standard Dynamic Range (SDR), which have a constant perceptual response, PQ is a non-linear function meticulously designed to match the human visual system's contrast sensitivity across an extremely wide luminance range—from 0.0001 to 10,000 nits.

The PQ curve works by allocating more digital code values (bits) to luminance levels where the human eye can perceive finer differences in brightness and fewer bits to levels where differences are less noticeable. This is based on Barten's model of visual perception, which accounts for factors like spatial frequency and background luminance. Technically, the function is defined such that equal steps in code value correspond to equal steps in perceived visual contrast (just-noticeable differences or JNDs) across the entire brightness range. When a video encoder quantizes linear light scene-referred data, it applies the inverse PQ function (the OETF) to create the digital signal. The display then applies the PQ EOTF to this signal to reconstruct light that is perceptually uniform to the viewer.

In the 3GPP architecture for media services (e.g., Multimedia Broadcast Multicast Service - MBMS, or Dynamic Adaptive Streaming over HTTP - DASH), PQ is a key component of the end-to-end video pipeline. It is specified within codec profiles and levels (e.g., for HEVC and later VVC) and in media presentation descriptions. Key components include the mastering display metadata (MaxCLL, MaxFALL) that defines the originating HDR monitor's capabilities and the signaling of the transfer function itself (e.g., 'pq' in the Colour Description Box of ISOBMFF files). Its role is to enable the efficient and high-quality delivery of HDR video content over mobile networks, ensuring that the breathtaking contrast and color volume intended by content creators are preserved from production through transmission to consumption on compatible HDR displays, such as those supporting HDR10.

Purpose & Motivation

PQ was created to overcome the severe limitations of the gamma curve (BT.1886) used for SDR television and early digital video. The gamma curve was designed for CRT phosphors and a limited brightness range (around 100 nits). It performs poorly for HDR because it wastes bits in the highlights and shadows where perceptual sensitivity is low and does not provide enough gradation in the mid-tones where the eye is most sensitive when viewing bright content. As display technology advanced to support thousands of nits, a new transfer function was needed to efficiently encode this extended luminance range within the bit-depth constraints of video codecs (typically 10 or 12 bits).

The problem PQ solves is bit-efficient perceptual encoding for HDR. Without PQ, representing a 0-10,000 nit range linearly would require an impractically high bit depth. PQ allows this vast range to be mapped into a 10 or 12-bit signal without visible banding or loss of detail in perceptually critical areas. It was motivated by the industry's move towards HDR video for premium content (movies, streaming) to provide a more immersive and realistic viewing experience with specular highlights, deep blacks, and vibrant colors. 3GPP adopted PQ (and later Hybrid Log-Gamma - HLG) to standardize HDR video delivery over mobile networks, enabling services like 4K/HDR streaming and broadcast, which are key differentiators for 5G multimedia services.

Key Features

• Non-linear EOTF based on human visual contrast sensitivity (Barten's model)
• Supports luminance range from 0.0001 to 10,000 nits
• Enables efficient 10-bit or 12-bit encoding of HDR video
• Standardized as SMPTE ST 2084 and adopted in 3GPP TS 26.116
• Used in HDR10 and Dolby Vision profile 5 media formats
• Requires associated static metadata (Mastering Display Colour Volume)

Evolution Across Releases

Defining Specifications