Catálogo de publicaciones - libros

Providing VOD in the internet is one of the challenging technologies. When a new client joins an ongoing multicast session for VOD service, the servers using CIWP scheme for the VOD multicast creates an additional unicast channel to serve the partial stream. And, the unicast channel consumes a certain amount of the I/O bandwidth of the server, as well as some of the network resources between the server and clients. This problem can be solved by using p2p local transfer between the clients to deliver the partial stream. In this paper, we propose a new VOD multicast scheme that is based on the CIWP scheme and the p2p transfer of the partial multimedia stream. In the p2p approach, unexpected dropout of a client, due to the failure of the connection or departure from the current session, can disrupt the partial stream transfer for the other clients. Thus, we also propose a procedure to recover from this kind of unexpected dropout. Our simulation results show that the proposed scheme reduces the network bandwidth on the server side dramatically, reduces the average waiting time of the client, and improves the service quality.

Various computing environments and user contexts require customized applications including composite services. We propose a middleware called SON-CMC (Service Overlay Network for Context-aware Mobile Computing) which supports multimedia service composition on overlay network for mobile computing. Our approach provides customized applications to mobile users by composing services such as web services based on user requirement and the contexts of mobile devices. Utilizing the large variation in quality of service between media servers and mobile devices, SON-CMC reconfigures the service composition graph to reduce the total data traffic and the response time of applications. It also aims to achieve load balancing by applying a service routing algorithm to the overlay network, which is modeled as a common peer-to-peer network. Experimental results demonstrate that our context-based scheme enhances service composition in terms of response time and preservation of network resources.

Peer-to-peer (P2P) based video-on-demand (VoD) systems rely on the cooperation among peers to reduce the server workload. Recently, hard cache is used to further improve the system scalability, because the contents will not be immediately cleaned up when the users get offline. However, how many practical benefits hard cache will bring to the P2P based VoD service has not been well studied and still remains far from clear. In this paper, we first characterize user behavior model with the benefit of millions of real VoD traces and identify several practical factors which potentially impact the system performance. Then we further conduct extensive trace-driven simulations to evaluate the scalability of P2P based VoD system with hard cache enabled and some interesting results are found.

In this paper, we described a framework for peer-assisted multi-path video distribution combined with cooperative caching. The target of this framework is to aggregate peers’ storage and bandwidths to facilitate video-on-demand streaming. To achieve this goal, we employ segment-based video caching and the segments are distributed in respective peers. Specifically, the source bit stream is based on layered scalable video coding for cost-effective video distribution. For achieving low -cost collaboration, a utility-based partial caching scheme is proposed and detailed discussed. Extensive simulations on large, Internet-like topologies were performed to demonstrate the effectiveness of this proposed framework.

Digital photography generates a lot more “shoeboxes” of photos than its conventional counterpart, resulting in image search and retrieval being more applicable. We briefly discuss some research challenges faced with the use of metadata in image search and retrieval. We then propose the structural use of metadata regularity of photos within collections (the ), in metadata management, reuse, inference and propagation. This application of the is complemented by the whereby user interactions with image collections provide collaborative metadata. This is followed by our presentation of a set-theoretic approach to our framework (proposed in previous work [5,6]) and we then outline its application and utility.

Multimedia surveillance systems utilize multiple correlated media streams, each of which has a different confidence level in accomplishing various surveillance tasks. For example, the system designer may have a higher confidence in the video stream compared to the audio stream for detecting humans running events. The confidence level of streams is usually precomputed based on their past accuracy. This traditional approach is cumbersome especially when we add a new stream in the system without the knowledge of its past history. This paper proposes a novel method which dynamically computes the confidence level of new streams based on their agreement/disagreement with the already trusted streams. The preliminary experimental results show the utility of our method.

Content-based cross-media retrieval is a new category of retrieval methods by which the modality of query examples and the returned results need not to be the same, for example, users may query images by an example of audio and vice versa. Multimedia Document (MMD) is a set of media objects that are of different modalities but carry the same semantics. In this paper, a graph based approach is proposed to achieve the content-based cross-media retrieval and MMD retrieval. Positive and negative examples of relevance feedback are used differently to boost the retrieval performance and experiments show that the proposed methods are very effective.

More and more applications require peer-to-peer (P2P) systems to support complex queries over multi-dimensional data. The retrieval facilities of most P2P systems are limited to queries based on a unique identifier or a small set of keywords. The techniques used for this purpose are hardly applicable for content-based video retrieval in a P2P network (CBP2PVR). In this paper, we present the design of a distributed P2P video sharing system that supports content-based video retrieval. First we will propose the compact signature generation of video shot which can be distributed in a P2P network and used as the basis for a source selection. Second, a Global Indexing structure based on proposed novel PVR-tree index schema allows communicating only with a small fraction of all peers during query processing without deteriorating the result quality significantly. We will also present experimental results confirming our approach.

Decision tree (DT) has great potential in image semantic learning due to its simplicity in implementation and its robustness to incomplete and noisy data. Decision tree learning naturally requires the input attributes to be nominal (discrete). However, proper discretization of continuous-valued image features is a difficult task. In this paper, we present a decision tree based image semantic learning method, which avoids the difficult image feature discretization problem by making use of semantic template (ST) defined for each concept in our database. A ST is the representative feature of a concept, generated from the low-level features of a collection of sample regions. Experimental results on real-world images confirm the promising performance of the proposed method in image semantic learning.

Classification is a major task in many applications and in particular for automatic semantic-based video content indexing and retrieval. In this paper, we focus on the challenging task of classifier output fusion. It is a necessary step to efficiently estimate the semantic content of video shots from multiple cues. We propose to fuse the numeric information provided by multiple classifiers in the framework of evidence logic. For this purpose, an improved version of RBF network based on Evidence Theory (NN-ET) is proposed. Experiments are conducted in the framework of TrecVid high level feature extraction task that consists of ordering shots with respect to their relevance to a given semantic class.