Research On Robust Video Commercial Detection Technique

Nowadays, multimedia data has permeated into our daily-life. In the face of such enormous multimedia information, how to get the right information we want expeditiously and precisely still keeps a challenging task. In order to achieve this objective, Content-Based Multimedia Information Retrieval (CMIR) has been studied as a new integrated application referring to many subjects and theories. Besides, as the main medium of commercial information during digital world, now commercial video information plays a more and more important role in transfers of commerce information. However, the research on advertisement detection still hangs behind compared with other detections. It is because that it is complex for computers to understand the advertisements. On the one hand, the producing and expressing skills are complicated and diverse without uniform rules. On the other hand, as the component of TV programs, it has to complete information during shot time with shot periods of characteristics. So it's difficult to detect. Considering the characteristics and structures of video and audio information during commercial programs, this paper proposes the robust video commercial detection system to detect commercial videos. And the research can be summarized as follows,â‘  In the stage of commercial shot segmentation, in order to segment TV programs into shots accurately, two Important Region Feature-Based schemes are proposed to get hard cut and dissolve cut shots respectively. And during the detection of cut shots, we apply the self-adaptive threshold to improve the robustness of our system.â‘¡ In the stage of merging commercial visual and audio information, we detect the audio changes in programs. With the characteristics of audio cuts, we refine the boundaries of all detected shots for getting the accurate commercial shots.â‘¢ During the post-processing, considering of the fusional continuance of time and content of commercial shots, we use a sliding window to refine the classification of shots in order to eliminate the wrong-classified shots. At last, commercial shots are merged to get commercial sequences by continuance of commercial time.â‘£ Based on the research above, we have set up a robust commercial detection prototype system.