Self-contained slices in H.264 for partial video decoding targeting 3D light-field displays

Abstract

Some video applications work with parallel bitstreams however only parts of them are required. Decoding the streams entirely might be prohibitive due to high number of video streams and / or high resolution or bitrate of the streams. In such cases, a random access also referred as to partial decoding has to be supported. It can be achieved with a conforming decoder, if the video is separated into individual slices, and the unnecessary slices are dropped from the bit stream. Such a feature is essential in 3D lightfield displays in which partial decoding of bitstreams helps providing less complex and thus real-time decoding and processing. This paper studies the problem of creating self-contained (i.e., independently decodable) slices in an H.264 stream. The requirements that need to be fulfilled in order to build selfcontained slices are described, and an encoder-side solution is proposed. The advantages of self-contained slices are examined in the context of a light-field rendering application that relies on partial decoding.