| Abstract |
Dear Readers, This issue consists of seven regular papers that were carefully peer-reviewed and selected by the editorial board. The published papers range from experimental investigations of Jini /Java implementation for multimedia control and management, to visual attention models, to caching Internet streaming media. I will briefly introduce each paper to provide you with a quick overview of the issue’s content. The paper titled “The design and implementation of a Jini/Java-based A/V streaming framework,” by C. Zhang, K.-Y. Koh, G.-M. Ong, and H.-K. Pung, reports on the design and implementation issues of new Jini/Java support for the OMG A/V stream specification and multipoint-to-multipoint streaming framework. This framework is accompanied by a receiver-based group communication mechanism that extends the semantics of membership to the leaving and joining of groups. With Java implementation, the performance is always of crucial importance. The authors report superior performance of their framework in comparison to Java RMI mechanisms; however, they report less favorable performance compared to UDP-based implementation or CORBA/C or C + + due to the higher processing overhead of Java. In the paper “Late join algorithm for distributed interactive applications,” authors J. Vogel, M. Mauve, V. Hilt, and W. Effelsberg address the support of participants who arrive late and wish to join an ongoing session of a distributed interactive application such as shared whiteboards and multiplayer computer games. The presented solution consists of an integration of a distributed late join algorithm and group communication. The authors’ solution is scalable, robust, and flexible as well as efficient in terms of three design criteria: low initialization delay, low network load, and low application load. Extensive simulation analysis, with a consideration of shared whiteboard and online gaming scenarios, validates the designed properties of the presented solution. All multimedia applications, whether interactive or on-demand, strongly rely on a sound multimedia file system that allows access to multimedia data with heterogeneous characteristics. Authors P. Shenoy, P. Goyal, S. Rao, and H.M. Vin present their experiences with the Symphony file system in the paper “Design considerations for the Symphony-integrated multimedia file system.” The paper provides a careful discussion of various methodologies for designing multimedia file systems and their trade-offs. Furthermore, the authors argue that to efficiently manage heterogeneity in multimedia application requirements and diverse data characteristics, a multimedia file system should enable the coexistence of multiple data-type-specific and application-specific techniques. They back up their claims through Symphony experiments in which their file system meets real-time requirements of video clients while improving by 1.9 times the text response time over conventional file systems. Image adaptation in adaptive content delivery for universal access has been actively explored for some time. Existing solutions generally do not pay much attention to user perceptions of the adapted result. The paper “A visual attention model for adapting images on small displays,” by L.-Q. Chen, X. Xie, X. Fan, W.-Y. Ma, H.-J. Zhang, and H.-Q. Zhou, presents a novel method of adapting images based on user attention. This method is based on three attributes: region of interest, attention value, and minimal perceptible size, associated with each attention object. The authors’ solution uses these attributes and dynamically modifies the image content (attention object) to optimally match the various screen sizes of client devices based on modeling of viewer attention. The method is implemented within an adaptive image browser. This browser provides not only the adapted view of important regions but also cropped parts of the original image via scrolling, assisting users to view the entire image. Experimental results demonstrate the feasibility and efficiency of the presented method. In most multimedia application domains such as entertainment and education, there is a strict constraint that the playback rate be the same as the capture rate. However, in the field of scientific and engineering analysis, it is often more appropriate to playback at a rate different than the capture rate. For example, automobile impact tests typically record multiple videos at 1000 fps but are played back at the rate of 25-30 fps. On the other hand, plant growth is captured at the rate of 1 frame per hour and played back at 25-30 fps. The paper “Synchronized multimedia for engineering and scientific analysis,” by Ch.S. Athwal and J. Robison, investigates this timing behavior in engineering tests and scientific experiments where a variety of techniques record phenomena not correlated to the human visual and cognitive time frame. The authors present synchronization methods to allow for variable start times and variable time increments both within one media stream and across a range of media elements. The successful validation of the presented techniques is done within a system for the analysis of vehicle crash tests. Automatic video shot motion characterization is an important function in video indexing and retrieval once the video is temporally segmented into shots. In the paper “Video shot motion characterization based on hierarchical overlapped growing neural gas networks” X. Cao and P.N. Suganthan present a neural-network-based scheme to characterize video shots by camera motion. The classification of different camera motions into shots uses a hierarchical overlapped incremental neural network model, and the architecture is based on the Fritzke’s Growing Neural Gas (GNG) network. The authors combine the unsupervised and supervised learning schemes in GNG, and the final classification is obtained by fusing individual classifications generated by the top-level overlapped GNGs. Using MPEG frames, experiments show a good classification accuracy. Access to streaming media requires a high and stable transmission rate. To meet such requirements, customers and ISPs typically upgrade the “last mile” connection to the Internet, for example, by going to higher-bandwidth services. However, this upgrade does not translate into improved quality of service for streaming media access. This and other factors suggest the importance of an efficient robust streaming delivery mechanism, such as caching, that goes beyond the point-to-point, server-client delivery of media content. S. Jin, A. Bestavros, and A. Iyengar in “Network-aware partial caching for Internet streaming media” present a novel technique that turns edge caches into accelerators of streaming media delivery. The cache management algorithms are stream-aware and network-aware. Using synthetic-generated access workloads, relying on models reflecting Internet bandwidth characteristics, and coming from real proxy cache logs, the simulations show that the cache management algorithms efficiently utilize cache space to reduce streaming media service delay and improve stream quality. The editorial board hopes that you will find the published regular papers interesting. |
