Decoded Picture Buffer

Description

The Decoded Picture Buffer (DPB) is a critical memory management component within video encoding and decoding processes standardized by bodies like ITU-T and ISO/IEC, and referenced in 3GPP for multimedia services. Its primary function is to store decoded pictures (frames) that have been reconstructed by the decoder. These stored pictures serve as reference frames for decoding subsequent inter-coded pictures (P-frames and B-frames). Inter-picture prediction relies on motion estimation and compensation, where blocks in the current frame are predicted from blocks in previously decoded reference frames stored in the DPB. The decoder uses the DPB to manage the availability and ordering of these reference pictures.

Architecturally, the DPB is managed by the Hypothetical Reference Decoder (HRD) model, which defines buffer management rules to prevent overflow and underflow, ensuring a consistent decoding rate. The DPB has a defined maximum size, specified in the bitstream, which constrains how many reference frames can be stored simultaneously. The decoder operations involve placing a newly decoded picture into the DPB, marking pictures as 'used for reference' or 'unused for reference', and removing pictures based on commands from the bitstream (like memory management control operations) or when they are no longer needed for prediction or output. The output process, which delivers pictures for display, is separate from the reference picture management, and the DPB ensures pictures are output in the correct display order, which may differ from the decoding order, especially with B-frames.

In 3GPP, the DPB is relevant within specifications like TS 26.119 and 26.906 that deal with multimedia codec conformance and performance for services like Multimedia Broadcast/Multicast Service (MBMS) and Enhanced Voice Services (EVS). The proper functioning of the DPB directly impacts video quality, bandwidth efficiency, and decoder complexity. If the DPB management is flawed, it can lead to visual artifacts, decoding errors, or buffer violations that break the decoding process. Therefore, 3GPP codec specifications include constraints and tests to ensure DPB behavior is predictable and interoperable across different devices and networks, which is essential for reliable video streaming and communication services.

Purpose & Motivation

The Decoded Picture Buffer exists to solve the fundamental problem of temporal redundancy in video sequences. Video compression standards achieve high compression ratios by not transmitting full frames repeatedly but instead sending differences (residuals) relative to previously transmitted reference frames. The DPB provides the necessary memory to hold these reference frames at the decoder, enabling motion-compensated prediction. Without a managed buffer for reference pictures, efficient inter-frame coding would be impossible, leading to drastically larger file sizes or bitrates unsuitable for mobile networks.

The historical context stems from the evolution of video coding standards from MPEG-2 to H.264/AVC and H.265/HEVC, each introducing more sophisticated DPB management to support features like multiple reference frames, adaptive reference picture selection, and hierarchical coding structures. In mobile communications, bandwidth is a precious resource, and video is a dominant traffic type. 3GPP's inclusion of DPB-related parameters and conformance points in its specifications (e.g., for MBMS or packet-switched streaming) ensures that video services are delivered efficiently and reliably across diverse devices. It addresses the limitation of simpler buffering models that could not support the advanced prediction tools required for high-quality low-bitrate video, which is essential for user experience on bandwidth-constrained cellular networks.

Key Features

• Stores decoded reference pictures for motion-compensated prediction
• Managed according to the Hypothetical Reference Decoder (HRD) model
• Has a defined maximum size to constrain decoder memory
• Supports marking pictures as 'used for reference' or 'unused for reference'
• Facilitates output reordering for correct display sequence (e.g., for B-pictures)
• Critical for conformance testing of video decoders in 3GPP multimedia specs

Evolution Across Releases

Defining Specifications