2.9 MPEG-7 Supporting Tools and Reference Software

< Day Day Up >

There are several tools available that support the annotation process with MPEG-7. The most complete one is the MPEG-7 Reference Software. The MPEG-7 Reference Software collected tools created during the MPEG-7 standardization process for most of the descriptors and description schemes and became part of the standard (Part 6). However, many of the tools generate XML fragments that conform only to an older version of the standard; others provide only a very limited functionality of what is proposed in the conceptual model associated to an MPEG-7 type. However, the reference software is a good starting point for any indexing and retrieval software relying on MPEG-7. Its current version can be downloaded at http://www.lis.e-technik.tu-muenchen.de/research/bv/topics/mmdb/e_mpeg7.html. This site also provides general information on the software. The reference software reached a stable version in March 2002. Many people have added their own implementations to it, based on the open framework proposed for descriptors and description schemes. Interested readers may follow the discussions on the mp7if_news group (http://groups.yahoo.com/group/mp7a_news/) set up by the MPEG-7 Alliance (see http://www.mpeg-industry.com).

Other tools support the annotation process by automatic extraction of visual features and by automatic segmentation of images and videos. Some of them propose semiautomatic creation of MPEG-7 documents, thus providing automatic low-level feature detection combined with a user-friendly interface for entering values for supplemental descriptors or description schemes.

Three commercial tools are of practical relevance. First, there is the IBM Annotation Tool (http://www.alphaworks.ibm.com/tech/videoannex), which assists in annotating video sequences with MPEG-7. Each shot in the video sequence can be annotated with static scene descriptions, key object descriptions, event descriptions, and other lexicon sets. The annotated descriptions are associated with each video shot and are stored as MPEG-7 descriptions in an XML file. It also allows customized lexicons to be created, saved, downloaded, and updated. The IBM Annotation Tool has a user-friendly interface consisting of all the necessary features for segmenting a video and classifying its content because of its lexicon, and is available on the basis of the AlphaWorks 90-Day Trial License Agreement. In addition to the IBM Annotation Tool, which proposes means for automatic segmentation, the sister tool from IBM (Multimodal Annotation [http://www.alphaworks.ibm.com/tech/multimodalannotation]) supports manual annotation of video and audio files on the basis of segmentation information obtained, for instance, from the Annotation Tool.

Second, the Ricoh MovieTool (http://www.ricoh.co.jp/src/multimedia/MovieTool/) is a tool for creating video content descriptions conforming to MPEG-7 syntax interactively. The MovieTool generates MPEG-7 descriptions based on the structure of a video. By using MovieTool, the user can create the structure while watching the video. Alternatively, MovieTool's editing functions can be used to edit the XML file, which contains the MPEG-7 description. A major advantage of using MovieTool is that the user can quickly and easily see the correspondence between the MPEG-7 descriptions and the video structure of each scene. Unfortunately, no trial version is available. The use of MovieTool requires one to agree to the terms and conditions of the support service contract, with an annual fee.

Although MovieTool generates MPEG-7 documents conformant to the standard, redundancy in the documents is high. For instance, for our pisa.mpg video, twice as many of the segments are described than are detected. These segments are dummies and do not have any useful description.

Third, the MPEG-7 Audio SpokenContent Tool by Canon generates a description file from an audio file in the wav format. The software is based on speech recognition technology being developed by Canon. It can only recognize phonemes (no words) and can only process speech signal files of limited size. Obviously, speech-recognition results depend heavily on acoustic conditions, individual speakers, and so forth. The authors tested their engine using Wall Street Journal data and achieved phoneme recognition rates of over 60 percent accuracy on the most likely phoneme string. More information may be obtained from http://www.cre.canon.co.uk/mpeg7asr/. Registration is obligatory for download of the software.

The Whisper Lab at the University of Wollongong developed a Web interface that runs an MPEG-7 audio low-level descriptors calculator. It calculates MPEG-7 descriptors from an audio file that is either in wav or MP3 file format and that is not longer than 1 MB. The online interface can be found at http://www.whisper.elec.uow.edu.au/mpeg7/.

Tools providing means for semantic annotation (i.e., for event, object, etc., descriptors) are actually under consideration. The actual list of available commercial tools may be found at the MPEG-7 Industrial Page at http://mpeg-industry.com/.

Recently, MPEG initiated two new MPEG-7 parts. Part 9 specifies profiles of description tools. The distinguishing feature is complexity. Part 10 specifies the schema definition across the parts of ISO/IEC 15938 and will contain a single schema across different versions. Parts 9 and 10 will be International Standard in 2004.

< Day Day Up >