ABSTRACT
XML-based metadata for digital media is becoming increasingly important,
as a consequence also calling for efficient encoding and compression
schemes for the storage and transport of this metadata. Moreover, support
for streaming the XML metadata in conjunction with the media data is
highly desirable. Such support is provided, for instance, by MPEG's
Binary Format for Metadata (BiM) encoding approach, which facilitates
fragmenting, delivering, and accessing the metadata in so-called Access
Units (AUs). In this paper, we present a quantitative evaluation of
existing XML metadata compression and encoding techniques, reaching from
widely used state-of-the-art data compression algorithms to sophisticated
XML-aware encoding schemes. The comparison is based on compressing MPEG-21
generic Bitstream Syntax Descriptions (gBSDs) which can grow to
non-negligible sizes. The main conclusion from this investigation is that
in terms of pure compression efficiency on XML files, the BiM approach
(exemplified by the MPEG reference software as well as a commercial version
thereof) is comparable – in terms of performance – with traditional data
or specific XML compression tools. However, when XML metadata have to be
fragmented, compressed, and streamed in such fragments, the results
indicate that the BiM approach is superior to the other schemes.