M.R. Naphade, T. Kristjansson, B. Frey, and T.S. Huang, "Probabilistic multimedia objects (multijects): a novel approach to video indexing and retrieval," in IEEE Interna- tional Conference on Image Processing, 1998, vol. 3, pp. 536--540.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....to users. The major disadvantage, however, is that many of these methods are heavily dependent on specific artifacts such as editing patterns in the broadcast programs, which makes them difficult to extend for the detection of other events. A more general method for the detection of events [17] uses Multijects that are composed of sequences of low level features of multiple modalities, such as audio, video, and text. Query by sketch or query by example methods have also been proposed recently [7] 36] to detect motion events. The advantage of these methods is that they are domain ....

....required. We believe the limitation to a specific domain, such as wildlife documentaries, does not limit our approach significantly, since such high level information is readily available from the content provider. The use of audio information represents one important difference to related work [17] that proposes a two level method using Multijects to combine low level feature information directly. Two other differences are: 1) the simplicity of the visual features they use to represent video frames and 2) their use of adaptive components (Hidden Markov Models) to learn the entire event ....

M. R. Naphade, T. Kristjansson, and T. S. Huang, "Probabilistic multimedia objects (MULTIJECTS): A novel approach to video indexing and retrieval in multimedia systems," Proc. IEEE Int. Conf. Image Processing, vol. 3, pp. 536--540, 1998.

....have generated a large body of knowledge and techniques. These techniques generally fall into one of the following research areas. Video indexing Research in this area aims at creating compact indices for large video databases and providing easy browsing and intelligent query mechanisms [3, 10, 33, 39]. Potential applications include multimedia databases, digital libraries, and web media portals. Video filtering and abstraction Research in this area tries to generate an abstract version of the video content that is important or interesting by extracting key portions of 4 the video [46, ....

....with slide transition information of the presentation, is used to extract important segments and generate summaries of oral slide presentations. This interesting application shows the strength of integrating audio features with other types of features, even meta information. Naphade et al. [39] used an integrated HMM model (called multiject) taking both audio and visual features as observations to detect events such as an explosion. Recently, they also proposed a probabilistic framework using Bayesian networks for semantic level indexing and retrieval [40] Sundaram and Chang [53] ....

M. R. Naphade, T. Kristjansson, B.J. Frey, and T.S. Huang, "Probabilistic multimedia objects (Multijects): a novel approach to video indexing and retrieval in multimedia 56 systems," Proc. IEEE International Conference on Image Processing, Vol. 5, Oct. 1998, Chicago, IL.

....and attribute values describing its content. In [CJ91] both the video objects and their spatial relationships are manually annotated in order to support complex spatial queries. In order to index video automatically, not only the visual contents but also the audio content should be used (see [NK98, NC97,WZ01]) Recent work also proposed the use of closed caption. An obvious trend in video indexing and retrieval is to use all the different features like audio, closed captions and so on instead of simply visual features. Another trend is towards web enabled based video management system. Over the last ....

M.R. Naphade, T. Kristjansson, B.J. Frey, and T.S. Huang, "Probabilistic Multimedia Objects Multijects: A novel Approach to Indexing and Retrieval in Multimedia Systems", Proc. IEEE Int' l Conference on Image Processing, Vol.3, pages 536-540, Oct 1998, Chicago, IL.

....belief networks, etc. The first publication addressing recognition of human actions using HMMs [12] describes the application of discrete HMMs in recognizing six different tennis stroke classes in a constrained test environment. Recently, similar techniques have been proposed. Naphade et al. [13] used hierarchical HMMs to extract events like explosions. HMMs together with a Bayesian classifier have been used for recognition of human actions in [14] Structuring of video using Bayesian Networks alone [15] or together with HMMs [16] has been proposed. In [17] a probabilistic model has been ....

M. Naphade, T. Kristjansson, B. Frey, T.S. Huang, "Probabilistic multimedia objects (multijects): A novel approach to indexing and retrieval in multimedia systems", In Proc. of the IEEE ICIP, Chicago, IL, 1998, vol. 3, pp. 536-540.

....with slide transition information of the presentation, is used to extract important segments and generate summaries of oral slide presentations. This interesting application shows the strength of integrating audio features with other types of features, even meta information. Naphade et al. [39] used an integrated HMM model (called multiject) taking both audio and visual features as observations to detect events such as an explosion. Recently, they also proposed a probabilistic framework using Bayesian networks for semantic level indexing and retrieval [40] Sundaram and Chang [53] ....

M. R. Naphade, T. Kristjansson, B.J. Frey, and T.S. Huang, "Probabilistic multimedia objects (Multijects): a novel approach to video indexing and retrieval in multimedia systems," Proc. IEEE International Conference on Image Processing, Vol. 5, Oct. 1998, Chicago, IL.

....to estimate the user s interests and to suitably modify the user s profile. We would also like to extend work in the domain of video. Video is an information intensive medium. Beside its temporal property, shared with audio, video has a spatial property which makes the problem more challenging [4, 60, 67]. For example, a user might request give me all video clips in which President Clinton and President Yeltsin are shaking hands. To respond to such a query we will need a data modeling technique which can support both spatial and temporal requests. We would like to consider cross modal queries ....

M. Naphade, and B. Frey, "Probabilistic Multimedia Objects (MULTIJECTS): A Novel Approach to Video Indexing and Retrieval in Multimedia Systems," in Proc. of IEEE Conference on Image Processing, Chicago, Oct 1998.

....fusion of multiple cues for shot boundary detection followed by the elimination results in a very robust high performance algorithm. Video segmentation paves way for efficient browsing and further analysis. It also paves way to developing algorithms for semantic video indexing search and retrieval [10]. Compact representation of video data can be done using key frames belonging to every shot. A simple way would be to represent every shot by its first frame. A more sophisticated approach is presented in [11] 5. CONCLUSIONS This paper proposes a new algorithm for shot boundary detection with ....

....management. We intend to use the shot boundary to analyze in greater depth and isolation, video data in each shot. We also intend to define shot similarity for efficient browsing and for query by example. Also a new approach is reported for semantic video indexing and retrieval is reported in [10], which makes use of this algorithm for shot boundary detection. The high performance shot boundary detection algorithm thus plays a central role in further video analysis. 6. ....

M. Naphade et al., "Probabilistic Multimedia Objects (Multijects) : A Novel approach to Indexing and Retrieval in Multimedia Systems," to be presented at the International Conference on Image Processing, Oct. 98.

....approach will evidently lead to far better results. Due to the inability for a universal definition of content, different types of content have been addressed, e.g. violence detection [3] scene change detection [4] news indexing [5] Limited efforts have been made to apply general descriptors [6] [8] Another approach attempts to provide visual or iconic summaries to the user to facilitate browsing [9] 11] A third approach aims at serving user queries by example [12] A content analysis method is presented which analyzes both auditory and visual sources and accounts for their ....

M.R. Naphade et al., "Probabilistic multimedia objects (multijects): A novel approach to video indexing and retrieval in multimedia systems", in Proc. of ICIP'98, 1998, vol. 3, pp. 536--540.

No context found.

M. Naphade, T. Kristjansson, B. Frey, and T. S. Huang, "Probabilistic multimedia objects (multijects): A novel approach to indexing and retrieval in multimedia systems," in Proceedings of the fifth IEEE International Conference on Image Processing, vol. 3, pp. 536--540, (Chicago, IL), Oct 1998.

....semantic retrieval for a small set of keywords and also act as the first step in QBE systems to narrow down the search. The di#culty lies in the gap between low level media features and highlevel semantics. Recent attempts to address this include detection of audio visual events like explosion [3] and semantic visual templates [4] We propose a statistical pattern recognition approach for training probabilistic multimedia objects (multijects) which map the high level concepts to low level audiovisual features. We also propose a probabilistic factor graph framework, which models the ....

....for the event explosion and site beach. User queries might similarly involve sky, helicopter, car chase etc. Detection of some of these concepts may be possible, while some others may not be directly observable. To support such queries, we proposed a probabilistic multimedia object (multiject) [3] as shown in Figure 1 (a) which has a semantic label and which summarizes a time sequence of features from multiple media. A Multiject can belong to any of the three categories: objects (car, man, helicopter ) sites (outdoor, beach) or events (explosion, man walking) Intuitively it is clear ....

[Article contains additional citation context not shown here]

M. Naphade, T. Kristjansson, B. Frey, and T. S. Huang, "Probabilistic multimedia objects (multijects): A novel approach to indexing and retrieval in multimedia systems, " in Proceedings of the fifth IEEE International Conference on Image Processing, vol. 3, Chicago, IL, Oct 1998, pp. 536--540.

No context found.

M. Naphade, T. Kristjansson, B. Frey, and T. S. Huang, "Probabilistic multimedia objects (multijects): A novel approach to indexing and retrieval in multimedia systems," in Proceedings of the fifth IEEE International Conference on Image Processing, vol. 3, pp. 536--540, (Chicago, IL), Oct 1998.

No context found.

M.R. Naphade, T. Kristjansson, B. Frey, and T.S. Huang, "Probabilistic multimedia objects (multijects): a novel approach to video indexing and retrieval," in IEEE Interna- tional Conference on Image Processing, 1998, vol. 3, pp. 536--540.

No context found.

M.R. Naphade, T. Kristjansson, B. Frey, and T.S. Huang, "Probabilistic multimedia objects (multijects): a novel approach to video indexing and retrieval," in IEEE Interna- tional Conference on Image Processing, 1998, vol. 3, pp. 536--540.

No context found.

M. Naphade et al., "Probabilistic Multimedia Objects (Multijects) : A Novel approach to Indexing and Retrieval in Multimedia Systems" to be presented at the International Conference on Image Processing, Oct 98

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC