Roger Midtstraum

Searching and browsing a shared video database

VideoSTAR (Video STorage And Retrieval) is an experimental video database system that gives support for sharing and reuse of video and meta-data. We discuss the need for a user to control the degree of meta-data sharing when searching and browsing a shared video database and we propose a way of organising meta-data to allow such functionality. This paper also presents a video query algebra utilizing the way meta-data are organised that allows the user to formulate temporal queries on the video contents. In the last part of the paper we discuss the use of the proposed concepts and the query algebra for video querying and browsing. We present a prototype video query tool implemented on top of the VideoSTAR framework that allows users to formulate quite complex queries based on the meta-data available in the VideoSTAR repositories.

A Temporal Foundation of Video Databases

Audio and video data represent streams of data with inherent temporal properties. In this paper we consider a video database as a collection of partial ordered sets where temporal relationships exist between elements from the same video stream. Video production introduces dependencies between different time coordinate systems. In this paper we give a formal definition of the contents of a video database. We also define mapped video object sets and operations on such sets that can be used for querying the temporal properties of video data and that can be used for mapping video objects between different time coordinate systems. At last, we illustrate how the proposed foundation can be used in searching and browsing.

Video server on an ATM connected cluster of workstations

Video servers are important for applications which make use of digital video. The video servers should provide better functionality than most of todays video servers offer,-e.g., support of flexible and instant user interactions, delivery of multiple video formats and support of virtual video documents. In this paper we discuss the requirements that video servers should fulfill and we describe the design and implementation of the Elvira video server. The Elvira video server is built on a cluster of standard UNIX workstations interconnected by an ATM switch. The capacity of the Elvira server is evaluated and we show the effects of different strategies for allocation of video delta across nodes and disks.

Low-cost access time model for serpentine tape drives

When a serpentine tape drive is used as a slow random access device, the I/O performance can be substantially improved by clever re-ordering of the I/O requests. This kind of re-ordering relies on a scheduling algorithm and a model of the access time. In this paper, we propose a low-cost access time model for serpentine tape drives, which is not a trivial task due to the complex data layout of serpentine tape. This model provides a way to estimate the physical positions on the tape for any logical data block, provides cost functions to estimate the seek time between two physical tape positions, and computes the transfer time of a data request. Our experiments show that the mapping from logical address to physical position has to be instrumented once for each tape cartridge. Algorithms are given to do this at a low cost. The accuracy of the model is assessed by measurements on rape drives and by use in scheduling of I/O requests. Experiments show that the model estimates are good enough to facilitate efficient scheduling of I/O requests.

Improving the access time performance of serpentine tape drives

The paper presents a general model for estimating access times of serpentine tape drives. The model is used to schedule I/O requests in order to minimize the total access time. We propose a new scheduling algorithm, Multi-Pass Scan Star (MPScan*), which makes good utilization of the streaming capability of the tape drive and avoids the pitfalls of naive multi-pass scan algorithms and greedy algorithms like Shortest Locate Time First. The performance of several scheduling algorithms have been simulated for problem sizes up to 2048 concurrent I/O requests. For scheduling of two to 1000 I/O requests, MPScan* gives equal or better results than any other algorithm, and provides up to 85 percent reduction of the total access time. All results have been validated by extensive experiments on Tandberg MLRI and Quantum DLT2000 drives.

Information models leveraging identification of returnable transport items

Purpose – New technology for equipping returnable transport items (RTI) with unique identification (e.g. RFID) give food industry companies new opportunities for improving business processes. The efficient exploitation poses new demands on the information technology systems in general and on the underlying information models in particular. This paper aims to introduce a generic information model where functionality for handling RTIs with unique identification is included.Design/methodology/approach – Functional requirements are analysed based on process descriptions and use case examples from a real production environment. Detailed information modelling was then done in several iterations using entity‐relationship diagrams. Validation of the model was done based on use cases typical for the food industry.Findings – The model described is developed within the broad scope of looking at opportunities for using wireless technology to improve business processes and traceability. This includes RFID, RF based se...

Databases for video information sharing

This paper describes the VideoSTAR experimental database system that is being designed to support video applications in sharing and reusing video data and meta-data. VideoSTAR provides four different repositories: for media files, virtual documents, video structures, and video annotations/user indexes. It also provides a generic video data model relating data in the different repositories to each other, and it offers a powerful application interface. VideoSTAR concepts have been evaluated by developing a number of experimental video tools, such as a video player, a video annotator, a video authoring tool, a video structure and contents browser, and a video query tool.

Software-Based copy protection for temporal media during dissemination and playback

We present a software-based protection mechanism to prevent unauthorized copying of media documents during their presentation on a clients host. Our solution enforces the execution and continuous replacement of security mechanisms on the clients host. Each protection mechanism is only used for a short time interval before it is replaced. The short duration of the time interval prevents a successful analysis and attack of the mechanism. In this way we solve a shortcoming of current solutions. As those employ fixed protection mechanisms they will eventually be circumvented because attackers have virtually unlimited time to analyze them.

SESAM: searching supported by analysis of metadata

This paper presents a novel approach for database searching where the user is assisted in locating relevant objects in query results. This is made possible by an active user interface which asks the user dynamically generated questions based on collected information of the properties of the objects in the result. The answers to these questions are used to retain only objects with the indicated properties, thereby improving precision. SESAM, a World Wide Web-based prototype based on this approach, is presented. It has been successfully tested using a large, real-life database.

Analysis of retrieval of multimedia data stored on magnetic tape

This paper discusses scheduling of random I/O requests for multimedia data stored on magnetic tape using serpentine data layout. Results from simulations and real experiments show that substantial improvements of retrieval performance can be achieved by proper I/O scheduling. A new algorithm, Multi-Pass Scan Star (MPScan*), is presented and shown to give better results than any other known algorithm. The level of quality of service that can be reached by use of MPScan* is discussed in the context of retrieval of image and short video segment data.