Reference Picture Selection Indication

Description

Reference Picture Selection Indication (RPSI) is a feature defined within the 3GPP Multimedia Telephony Service for IMS (MTSI) framework, specifically detailed in TS 26.922. It operates at the application layer as part of the video codec control protocol. The core function of RPSI is to provide a feedback mechanism from a video receiver to a video sender, instructing the sender to encode future video frames using a specific, older reference picture that the receiver has successfully decoded and stored in its buffer. This mechanism is triggered when the receiver detects that a current or recent video frame (or its dependent frames) has been lost or corrupted during transmission. By signaling the sender to switch to a known-good reference picture, the receiver can break the chain of error propagation that is typical in predictive video coding (like H.264/AVC or H.265/HEVC), where each frame often references the immediately preceding one. The RPSI message contains an identifier for the specific reference picture, allowing for precise recovery. Architecturally, it is part of the end-to-end control signaling between MTSI clients, independent of the underlying transport network's error correction mechanisms. Its role is crucial for maintaining acceptable video quality in conversational services where low latency is paramount and retransmissions are often not feasible.

Purpose & Motivation

RPSI was introduced to address the challenge of maintaining video quality in packet-switched real-time communication services like Voice over LTE (VoLTE) video calls. Traditional video codecs are highly efficient but vulnerable to packet loss; a single lost packet can corrupt a frame, and due to predictive coding, this corruption propagates to all subsequent frames that reference it, causing visible artifacts or a frozen picture until the next independently coded frame (I-frame) is received. In low-latency conversational services, waiting for a periodic I-frame is unacceptable. RPSI provides an immediate, application-layer recovery mechanism. It solves the problem by allowing the receiver to proactively guide the encoder's reference picture selection, enabling rapid recovery from errors without the bandwidth overhead of frequent I-frames or the latency of transport-layer retransmissions. Its creation was motivated by the need to make IMS-based video telephony robust over IP networks, which are inherently best-effort and prone to jitter and loss, especially in early LTE deployments and challenging radio conditions.

Key Features

• Application-layer feedback mechanism for video error control
• Signals a specific reference picture identifier from receiver to sender
• Breaks temporal error propagation in predictive video coding
• Operates within the 3GPP MTSI and IMS framework
• Reduces dependency on intra-frame refresh for error recovery
• Enhances user experience for real-time video calls over packet networks

Evolution Across Releases

Defining Specifications