Multimedia and imaging databases [Book Reviews] - IEEE Xplore

0 downloads 0 Views 377KB Size Report
and video); 2) enable new access meth- ... tutorial yet comprehensive fashion. This ... (OMG) and OLE 2.0 from Microsoft. ... query and retrieval in multimedia.
Building Neural Networks David M. Skapura, ACM Press, 286 pages, 1996, ISBN 0-201-53921-7

REVIEWER: SEUNGYUP PAEK The author of this book says that there is a big difference between knowing the theory associated with neural networks and actually applying neural networks to solve practical problems. Although there are many books available on the theory of neural networks, the author claims that there is a serious lack of books that provide guidance in developing applications using neural networks. Therefore, the main goal of this book is to fill this void and to provide insights into building neural network applications. T h e book starts with a chapter reviewing the fundamentals of neural networks, which is clearly written and easy to understand for readers with no experience in the area. This is followed by a chapter that reviews several well known neural network paradigms and tries to compare the practical aspects of the different neural network models. This chapter is a high-level overview and, as the author points out, beginners to neural networks will probably need a supplementary text in order to understand the workings of each neural network model. Following this chapter, the author has one chapter devoted to the process of actually building a neural network application. The author focuses on practical but important issues such as data formatting, network training, and performance evaluation. The chapter concludes with a useful detailed example of a practical neural network application. Finally, the next six chapters of the book each deal with different applications of neural networks such as image processing, pattern classification, robotics, and business applications. Each of these six chapters can be read independently, and each application uses the neural network models that were covered in the early portion of the book. Due to the broad scope of the applications and the limited size of the book, each chapter can only provide an introduction to each application area. As one example, the chapter on image processing does provide a good introduction to applications of neural networks to image processing. However, it does not provide enough details to actually build an application. Rather, it pro-

20

vides a good set of pointers and references to several of the major works in the area. Overall, the book is clearly written and does provide an interesting introduction to applying neural networks to practical applications. However, the book does require supplemental texts for a clearer understanding of the different neural network models. Perhaps the main contribution of the book are the pointers and references it provides to start work in different application areas of neural networks.

Multimedia and imaging Databases Serag Khoshafian and A. Brad Baker, Morgan Kaufmann, 586 pages, 1996, ISBN 1-55860-312-3

REVIEWER: CHUNG-SHENG LI The multimedia revolution during the past decade has already produced a profound impact on computing, business, and even our daily life. Multimedia information is space intensive. Multimedia elements need t o b e searched, concurrently accessed, updated, and manipulated consistently. Consequently, there is a clear and genuine need for a newer generation of database management systems which 1) allow the storage of various kinds of multimedia data types (time series, audio, images, and video); 2) enable new access methods, including content-based indexing; 3) enable automatic migration of data between different storage hierarchies; and 4) permit the concurrent access by a large number of users. Multimedia and Imaging Databases by Khoshafian and Baker is probably one of the first few books addressing all of these issues under a single cover, while keeping the whole discussion in a tutorial yet comprehensive fashion. This book is targeted towards professionals, managers, and high-level executives. But it is also an excellent book for any people who are interested in the emergence of multimedia networking, objectrelational database, object-oriented databases, multimedia databases, digital libraries, and hierarchical storage management systems. This book is divided into 12 chapters. The first chapter gives a brief introduction to the necessities of multimedia databases, the structure of clientherver architectures, and the integration of databases with the hierarchical storage management. Chapter 2

describes the graphical user interface, authoring mechanisms for multimedia systems, and various multimedia applications such as forms, compound document, hypermedia, and document imaging systems. T h e third chapter describes various multimedia peripherals and devices. These devices include monitors, scanners, sound cards, video boards, video cameras, printers, plotters, touch screen devices, storage devices, and so on. T h e interesting part of this book begins with the fourth chapter on object orientation for multimedia databases. The criteria for object-oriented database system such as abstract data types, inheritance, and object identity are discussed in detail. Examples for schema development and object-oriented database design are also given. This chapter certainly has a moderate bias toward object-oriented database systems. For a balanced view, the readers should also consult with the ObjectRelational DBMSs - The Next Great Wave by Prof. Stonebraker, which certainly has a strong bias toward the object-relational databases. Chapter 5 describes the clientisemer architecture and multimedia databases. The key issue discussed in this chapter is client-server relationships in a n object-oriented environment. An object responding to a message is acting as a server, the object sending the message is the client. This concept brings up two competing standards on component object computing: Common Object Request Broker Architecture (CORBA) from the Object Management Group (OMG) and OLE 2.0 from Microsoft. The rest of this chapter surveys several client/server examples including file servers, database servers, video servers, and fax servers. This chapter concludes with case studies on Oracle Media Server and IBM LAN server. Alternative strategies for incorporating multimedia support in relational and object-oriented databases are discussed in Chapter 6. The most critical components are mass storage and content-retrieval capabilities support in addition to traditional database support. This chapter also reviews language standardization trends (SQLS,ODMG), and multimedia support in commercial object-relational and object-oriented database. Different issues for dealing with query and retrieval in multimedia databases are described in Chapter 7 . A critical issue is the search and retrieval of any collection of multimedia objects. (Continued on page 24)

IEEE Communications Magazine * June 1998

(Continued from page 20) The indexing of multimedia documents and/or objects is either manual or automatic. Automatic indexing means index terms are algorithmically indexed and retrieved. Depending on the type, multimedia objects can have many attributes and features. Some of these features and attributes can be derived from the data (such as texture, color histogram, and shape), while the others are relevant to the multimedia objects but are usually captured through other mechanisms (such as author’s name, title, etc.). This chapter discusses many different techniques for content-based retrieval of text and images. Contentbased retrieval differs from conventional database retrieval both in terms of the query and indexing. Contentional database retrieval is usually precise, while content-based retrieval is mostly fuzzy. Consequently new interactive query paradigm such as relevance feedback, in which the user refine and resubmit the query, has to be developed for content-based queries. Chapter 8 focuses on some classical database issues which also need to be carefully considered in multimedia databases such as concurrency and recovery. Access methods, essential for efficient access of large database systems, are investigated in Chapter 9. Spatial indexing techniques such as K-d trees, grid files, quadtrees, and R-tree (multidimensional extension of B-trees) are discussed with some detail. As mentioned earlier, one of the key aspects in a large multimedia database system is the storage management subsystem. Consequently, an entire chapter (Chapter 10) is devoted to the optical file system, caching algorithms, and hierarchical storage management system (HSM) that are commonly associated with a large multimedia database systems. This chapter is concluded with case studies on several commercially available HSM systems. The communication and networking aspects of multimedia systems is discussed in Chapter 11. This chapter gives a very high-level overview of a number of recent networking technologies such as FDDI, high-speed Ethernet, frame relay, SONET, and ATM. Chapter 12 gives a summary of the book. The common pitfall shared by many books on multimedia systems is their ambitious goal to cover everything under the sun related to multimedia. This usually results in superficial treatment of each individual subject. Fortunately, this book has stuck to the main

24

focus and does give a very thorough and complete discussion of issues concerning multimedia databases. T o a large extent, the authors have also done a very good job in not lending any bias when discussing each individual subject. In summary, this book is really an outstanding book on a n important emerging area: multimedia databases. It is definitely strongly recommended for the professional working in this area and like to get an update, and is also highly recommended for students and researchers who are interested in starting working in this area.

Optical Networks Rajiv Ramaswami and Kumar N. Sivarjan Morgan Kaufmann Publishers, 1998,ISBN 1-55860-445-6, 607 pages

REVIEWER:CHUNG-SHENG LI In the past three years, we have witnessed a major change of heart on the optical networking technologies. Viewed as an obscure industry segment at its inception in the late ‘70s and early %Os, the whole area of optical networking has made slow progress (financially, not technologically) into t h e mainstream during its first decade. Mostly due to the recent explosive growth of Internet and Intranet traffic, the bandwidth required by telecommunications networks certainly outpaces fiber deployment. Consequently, there is an exacerbated demand for multiplexing multiple data streams over the same fiber e i t h e r through time-division multiplexing or wavelength-division multiplexing, This stronger demand is further evidenced by Wall Street’s recent strong interest in startup companies such as Ciena, which has supplier agreements with Sprint and Worldcom, and Positron Fiber Systems, which has strategic alliances with Digital Microwave and Siemens. Both these companies specialize in optical networking equipment and are projected to have better than 40 percent earningper-share growth rate for the next five years. As a result of the broadened audience, there is an increasing need for a good textbook that can give a self-contained introduction to both optical technologies and networking considerations for those students and professionals who want to have a jump start in this field. A number of books covering optical communications (most of them have been reviewed in this column in the past few years), such as Optical Fiber

Communications by John Senior (Prentice Hall, 1992), Optical Communicalzon Systems by John Gowar (Prentice Hall, 1993), and Fiber-optic Communication Systems by Govind Agrawal (Wiley, 1997) mostly focus on the physical layer aspects and do not really investigate networking issues. In contrast, Optical Communication Networks by Biswanath Mukherjee (McGraw Hill, 1997) focuses entirely on multi-access networking issues. Up to this point, only Fiber Optic Networks by Paul E. Green, Jr. (Prentice Hall, 1993), and, to a lesser extent, Fundamentals of Multiaccess Optical Fiber Networks by Denis J. G. Mestdagh (Artech House, 1995), have attempted to provide balanced and self-contained treatments of both physical layer and networking issues. Although each of the three approaches has its own rationale, the last approach is best for those people who need to acquire understanding of both areas simultaneously, because those issues have become increasingly correlated with each other. Consequently, it is no surprise that Fiber Optic Networks by Paul Green has been the benchmark in this area in the past few years. Optical Networks by Rajiv Ramaswami and Kumar Sivarjan is definitely a very timely update in this area. Both of the authors are highly qualified and have published in this area for almost a decade. Similar to Fiber Optic Networks, this book contains an excellent mix of both component technologies and networking infrastructures. The book is divided into two parts. The first part introduces basic subjects in fiber optics with emphasis on those issues that are strongly related t o multiaccess networks. This part includes component technologies for optical networks (e.g., optical amplifiers, transmitters, receivers, switches, wavelength converters, multiplexers, and filters), modulation and demodulation techniques, noise considerations, and transmission system engineering (crosstalk, dispersion, and fiber nonlinearities). These topics serve as solid foundations for understanding interesting (and sometimes peculiar) considerations in all optical networks where each fiber needs to carry more than one wavelength. T h e real focus of this book is, of course, on the second part - all-optical networks. It begins with a survey of the first-generation optical networks (mostly point-to-point) such as SONET, Fiber Channel, and HIPPI. Then, a great deal of this part is devoted to the investigation of broadcast-and-select and wavelength (Continued on page 26)

IEEE Communications Magazine June 1998