Enhancing Video Viewing Experience

Assisting People In Video Viewing

To support the video viewing process, numerous studies focusing on content-based analyses for retrieving or summarizing videos have been reported (Nakamura & Kanade, 1997; Ekin et al., 2003). These studies assume that content-based knowledge may include the semantic information of video data; in other words, it includes generally accepted domain knowledge. For example, people tend to pay additional attention to goal scenes during football games or captions during news programs that describe the summary or location of the news topic. Therefore, these approaches are applicable only for specific purposes (e.g., extracting goal scenes from football games as important scenes), which are assumed beforehand. However, it is difficult to determine the scenes of interest for a particular user. Some users might like to watch a summarized video, which focuses particularly on the goal scenes in a football game video, whereas other viewers might want to watch the scenes in which a particular star player appears. Further, in the case of a weather forecast video, users might want to watch the scenes showing the area in which they reside (Figure 1). Therefore, it is difficult for systems using content-based knowledge to satisfy the needs of users because the users’ intention or knowledge, which determines what they want to watch, is not entirely predictable, and occasionally, the users themselves cannot distinctly describe their viewing requirements.

Figure 1.

It is difficult for systems using content-based knowledge to satisfy the needs of users because the users’ intention or knowledge, which determines what they want to watch, is not entirely predictable, and occasionally, the users themselves cannot distinctly describe their viewing requirements