Reference Picture List Construction

Description

Reference Picture List Construction (RPLC) is a fundamental procedure in video compression standards adopted by 3GPP for multimedia services, such as those specified in TS 26.906 for Enhanced Voice Services (EVS) with video. It involves creating and managing ordered lists—typically List 0 and List 1—that contain reference pictures (previously decoded video frames) used for inter prediction during video encoding and decoding. Inter prediction reduces redundancy by encoding a current picture relative to reference pictures, using motion vectors to describe displacements. RPLC defines how these reference pictures are selected, ordered, and indexed, directly influencing coding efficiency, error robustness, and decoder complexity.

Architecturally, RPLC operates within the video codec's prediction framework, such as in H.264/AVC or H.265/HEVC codecs used in 3GPP systems. During encoding, the encoder decides which reference pictures to store in a Decoded Picture Buffer (DPB) and signals RPLC commands to the decoder via the bitstream. These commands instruct the decoder on how to construct identical reference picture lists, ensuring synchronization. The lists can include short-term and long-term reference pictures, with ordering based on picture order count, temporal distance, or other criteria. RPLC mechanisms handle operations like picture marking (as 'used for reference' or 'unused'), list modification, and default list initialization based on slice headers or sequence parameter sets.

In operation, RPLC is critical for both normal playback and error resilience. For example, in predictive video coding, a P-slice may use List 0 (for forward prediction) to reference past frames, while a B-slice may use both List 0 and List 1 (for bidirectional prediction) to reference past and future frames. The construction process determines which references are most relevant, affecting motion search efficiency and rate-distortion performance. In error-prone mobile environments, RPLC can be configured to enhance robustness—e.g., by prioritizing reference pictures that are more likely to be correctly received or by using redundant reference management. This helps conceal packet losses and maintain video quality under varying network conditions.

Purpose & Motivation

RPLC was standardized within 3GPP to optimize video transmission over wireless networks, where bandwidth is limited and channels are error-prone. Early video codecs had simpler reference picture management, which could lead to inefficiencies in compression or poor error recovery. RPLC provides a flexible, standardized method to construct reference lists, enabling better adaptation to content characteristics (e.g., motion, scene changes) and network conditions, thus improving overall quality of experience for multimedia services.

Historically, as mobile video usage grew, 3GPP incorporated advanced video codecs like H.264 into specifications such as EVS to support high-quality voice and video calls. RPLC addresses limitations of fixed reference picture schemes by allowing dynamic list construction, which enhances compression through more accurate motion prediction and supports features like hierarchical coding structures. This is particularly important for real-time communication, where low latency and high resilience are required.

Furthermore, RPLC facilitates error resilience techniques such as reference picture selection, where decoders can switch to alternative references if primary ones are lost. This reduces visual artifacts and call drops in volatile radio conditions. Its introduction in Release 12, as part of enhanced video codec profiles, reflects 3GPP's focus on improving multimedia performance in LTE and beyond, ensuring efficient use of spectrum and network resources while maintaining user satisfaction.

Key Features

• Constructs ordered lists of reference pictures for inter prediction in video coding
• Supports both forward (List 0) and bidirectional (List 0/List 1) prediction modes
• Enables dynamic management of short-term and long-term reference pictures
• Enhances compression efficiency through optimized reference selection
• Facilitates error resilience by allowing redundant or alternative reference pictures
• Signaled via bitstream commands to ensure encoder-decoder synchronization

Evolution Across Releases

Defining Specifications