Kazuhisa IGUCHI, Research Engineer, Advanced Audio & Video Coding I have been with NHK Science and Technical Research Laboratories since 1993. My research interest is image processing. I have been engaged in research on motion vector estimation and motion compensation for video standard conversion and object-based image processing. Since 1999, my research interest is scene description language. Existing scene description languages are principally for scene presentation to users. For reason of efficient program production, I think it is useful to apply scene description language to broadcast program production. If a standardized scene description language independent of specific hardware can be used in program production, it is possible for editors and studios to collaborate and easily modify or reuse existing image scenes.

The digitalization of broadcasting increases the availability of channels and the spread of the Internet is driving demand for a wider variety of programming. STRL is developing efficient program production technologies that can produce quality programs in a fraction the time and cost of conventional production systems. As part of this research, we developed a scene description language called media object interlock and synthesis (MIKSS for short and pronounced 'mix') that can describe the functions used in image composition, video effects creation, and editing. The development of this language was based on an examination of the functions needed for video production.
The objects comprising a video scene, such as people and the background, are called video components. In addition to indicating the display position and time for each component, MIKSS makes it easy to describe the video production functions. These functions include the video editing processes, using an "IN" point (the starting point of a video sequence) and an "OUT" point (the completion point of a video sequence), the scene change processes (switching of a video) such as wipe and dissolve (switching of a video by overlapping it with another), and the video effects processes, including color conversion and transformation. The language can also be used to describe interactive video scenes, providing additional video information whenever a user clicks a mouse or takes some other action.

Figure 1: Video composition by scene description

Figure 1 shows an example of video production using MIKSS. Moving images, such as the fish and water plants of the figure, are the video components. First, a scene description file containing the video-component display positions and times is created. This file is read by a scene description player, which displays a video scene containing the described video components. The video image can be easily modified, for example, replacing a video component or moving a display position, by rewriting the MIKSS scene description file.
Figure 2 shows a conceptual diagram of a video production system in which video components are shared on a network. Each area enclosed by broken lines comprises one editing system. The common scene description language allows a producer to edit on any editing system, as long as he or she has the scene description file. Since a scene description file can be created on a simplified editing system that handles only textual data, the costs for program production facilities can be reduced.
Future improvements to the language will include adding functions for video production via networks and making the video scenes described by MIKSS available for use in data broadcasting and on the Internet.

Figure 2: Video production system using scene description language