In this paper, we propose a method for two typical mid-level descriptors extraction of soccer video: view type and playfield position. The output reveals high-level structure of the soccer video. After the robust playfield segmentation, we use the grass-area-ratio and the boundary approximate lines between playfield and non-playfield to classify each frame into three kinds of view types according to a series of rules. We then classify frames captured by the main camera into fifteen kinds of positions by two steps: first, we classify the frames into five kinds according to the orientation of boundary lines; secondly, we use lines detection to further determine the specific playfield position. Experiments on three hours soccer videos show promising results.

There are usually repetitive sub-segments in broadcast videos, which may be associated with high-level concepts or events, e.g., news footage, repeated scores in basketball. Unsupervised mining techniques provide generic solutions to discovering such temporal patterns in various video genres, which are currently the subject of great interests to researchers working on multimedia content analysis. In this paper, we propose a novel approach to automatically detecting repetitive patterns in a video stream. In this approach, a video stream is first transformed to a symbol sequence via the spectral clustering algorithm. After computing the transition probabilities of any two symbols in temporal evolution, we produce a set of probabilistic templates to characterize the patterns of potential interest. Finally, we verify each probabilistic template by measuring the similarities between the video sub-segments and the template. Evaluations on various sports videos show promising results.

The motion vector field (MVF) represents motion characteristics in video sequences, and has been widely used and been proved to be effective in sports video analysis. However, in tennis video analysis, MVF-based methods are seldom utilized for the reasons that: (i) the players' true motion is not accurately represented by the extracted motion vector, due to the deformation caused by the diagonal shooting of the camera; and (ii) the motion vector's magnitude is not prominent enough and is prone to be disturbed by noises. In this paper, a pinhole camera modeling of the motion vector field is proposed to revise the deformed motion vector. In this modeling, a foreground object mask is adopted and global motion compensation is incorporated as the pre-processing steps. Evaluation of the proposed modeling using four hours of tennis videos shows very encouraging results.

A novel highlight browsing approach for sports videos, based on contextual cue categorization, is proposed in this paper. Contextual cues are those salient audio and visual cues around highlight segments, and they are also informative and attractive to audiences. With further analysis and categorization of these contextual cues, the user could access highlights more flexibly and effectively. An experimental system is built for broadcasting diving videos in our current work, in which contextual cues such as slow-motion replays and game statistics information are extracted and categorized for highlight browsing. An efficient algorithm for slow-motion replay detection and categorization in diving videos is also proposed. Experimental results on 2 hours broadcasting diving videos have demonstrated the effectiveness and efficiency of the proposed approach.

This paper presents a new approach to people-based video indexing. In this approach, we define a people-based similarity measure according to both clothing similarity and speaking voice similarity. Such similarity depicts how perceptually similar two people appearing in different scenes are and if they belong to an identical person. Instead of computing in feature space, the proposed people-based similarity is computed in distance space. The extended support vector machines (SVM) are employed to map a serial of low-level feature distances to a perceived people similarity. In order to build people-based video indexing, a novel unsupervised clustering algorithm is also proposed, which can more correctly identify an individual person according to mutual people similarities between two people. The experiments on large video testing data have demonstrated the effectiveness and efficiency of the proposed people-based similarity, unsupervised clustering and video indexing.