Fig. 1. The process of generating video proceedings.

Fig. 2. Three remote controlled pan/tilt/zoom cameras are installed from the ceiling center. A forth camera (not shown) is installed from the side of the auditorium. Courtesy of Ron Edmondson.

Fig. 3. The video production center provides full control of pan, tilt, zoom, focus, exposure, white balance, and other camera setting for each of the four cameras, live video editing and audio mixing. All the equipment is operated by a single person to produce high quality video. Courtesy of Ron Edmondson.

Fig. 4. This graph, captured from the SBD debugging program, represents 1200 frames (40 s). Color difference at 1, 3, 5, 7, and 12 frames apart are shown by graphs D1, D3, D5, D7, and D13. Graphs E1, Img, and Illum represent edges difference, thumbnail image difference, and frame average illumination, respectively. Graph Sys corresponds to the machine state during the process, and GT corresponds to two graphs, representing the algorithm output and the ground truth. Algorithm errors are marked in red color (see arrow). Twelve cuts and six dissolves are present in this window, all but this one are correctly detected.

Fig. 5. A slide show composed of timed JPEG images.

Fig. 6. A simple slide show composed of an RP file and an audio stream using the SMIL protocol.

Fig. 7. Left: synchronized media browser with multiple views. The buttons bellow the player allow to switch between views, to scroll through the search results, and to generate a URL link of the current point in the current video and mail it (see the popup mail window). Right: one of the views is a typical storyboard. Selecting a frame starts the player at the beginning of the corresponding shot.

Fig. 8. (A) Subjective comprehension decreases at speedup ratios higher than 1.5×. (B) Assessed comprehension decreases at speedup ratio higher than 1.7×. The three graph pairs correspond to different video domains: news (solid), training (dashed), and technical talk (dotted).

Shot boundary detection results, comparing the current system with the baseline system on both TREC-01 and TREC-02 video data test sets