Multi-view Video Coding

Description

Multi-view Video Coding (MVC) is a video compression standard developed as an amendment to H.264/Advanced Video Coding (AVC). Its primary function is to encode video sequences captured simultaneously from multiple cameras arranged to cover the same scene from different angles. The fundamental technical approach is hierarchical B-picture prediction extended across both time (within a single view) and view (across different camera perspectives). This creates a prediction structure that maximizes compression by using already encoded pictures from other views as references, in addition to past and future pictures in the same view.

The MVC codec architecture designates one view as the base view, which is encoded compliant with the H.264/AVC High Profile, ensuring that any AVC decoder can decode this view for 2D playback. Additional auxiliary views are encoded using inter-view prediction. Key coding tools include motion-compensated prediction for temporal redundancy and disparity-compensated prediction for spatial redundancy between views. The encoder generates a single bitstream containing all views, interleaved according to the prediction hierarchy. The bitstream includes Network Abstraction Layer (NAL) units with new types to identify view components and manage the decoding order.

Within 3GPP ecosystems, MVC has been specified for use in Packet-switched Streaming Service (PSS) and Multimedia Broadcast/Multicast Service (MBMS) for delivering 3D video content. A network service delivers the MVC bitstream, and a compatible client (e.g., a mobile phone with a 3D display) decodes at least two views to render a stereoscopic image. The 3GPP specifications define profiles and levels for MVC, its carriage in MPEG-2 Transport Streams or ISO Base Media File Format, and signaling in session description or media presentation descriptions. Its role was to provide a standardized, efficient method for 3D video delivery, forming the basis for early mobile 3D TV services.

Purpose & Motivation

MVC was created to enable practical 3D video services, which require at least two views (for left and right eye), at a time when transmitting two independent AVC streams would have doubled the bandwidth requirement. The primary problem it solved was the high bitrate cost of multi-view video, which was a major barrier to deploying 3DTV and other immersive services over bandwidth-constrained channels like mobile networks.

The historical context is the rise of 3D cinema and the initial push for 3D television in the late 2000s. 3GPP began incorporating support for advanced video services around Release 9. MVC, standardized earlier by MPEG, was adopted into 3GPP to provide a technically sound solution. It addressed the limitation of simulcast (independent streams) by providing typically 20-50% bitrate savings for two-view stereo content, making mobile 3D video broadcast and streaming commercially and technically viable. It laid the groundwork for all subsequent multi-view coding work in 3GPP, including the later MV-HEVC.