Browse Books

Reviewer: Ralph Raby

Despite its title, this book has little to do with the mechanics or technicalities of picture production in interactive systems, or even with information that can be subtly hidden within an image, such as digital watermarks. It has more to do with pictures in general—not just in interactive systems. The title might, therefore, be considered misleading, particularly since the word “pixels” applies equally to noninteractive graphics. Pixels form the basis of pretty much all modern computer graphics output, both on- and off-screen. The book is divided into six parts. Part 1 introduces some preliminary concepts of pictures and their relation to society, philosophy, and everything else. Part 2 introduces fundamental concepts related to pictures in general and their use in interactive systems. The authors classify pictures in three ways: abstract-graphical, pictograms, and presentational. Parts 3 through 5 treat each of these in turn. Part 6 presents concluding remarks. There is a comprehensive bibliography, and, in the spirit of the book, the index uses pictorial letters to mark out each alphabetical section. Many examples of the use and misuse of pictures are presented throughout, with particular reference to computer-generated pictures. For some applications, they may seem too perfect, and need, perhaps, an artistic brush to render them slightly less so. In other cases, scrappily drawn diagrams need to be made more perfect, using straight lines that meet at the corners, and so on. The authors note that there has been a recent trend toward too much photorealistic rendering and that photorealism does not suit all applications. They suggest that there is not enough scope for personal expression in modern computer graphics software and that many of the illustrations found in books might be difficult to generate exactly using computer graphics. I think that photorealistic images always look computer-generated, and that they are certainly not a substitute for real photographs (see the simulated faces of Bogart and Monroe). One important use of on-screen graphics is, through animation, to show a series of changes taking place over time. The computer can eliminate a large part of the drudgery associated with this task. The pictorial element involved caters to only one of the human senses. Other modalities, such as sound and touch, should be explored for enhancing information visualization. The authors devote a couple of chapters to tactile graphics and immersive virtual reality. Pictures and natural language are compared and contrasted, but the authors have almost lost sight of the fact that text itself is also a graphical element. Once integrated into other pictorial forms, it may lose its identity, in the sense of being unrecognizable to sentient computer software. A human viewer, of course, can easily pick out the text element and use it to understand the picture (What clever image-processing machines we are!). This book makes an interesting read, although some of it has been done before, and better, by authors such as Tufte. It is turgid and heavy in places and often over-repetitive. Ideas are introduced in one chapter and then dressed up slightly differently later on, as though some fundamentally new theory is being expounded. Much of what is presented seems to stem from common knowledge or common sense, but is spelled out at length. The “transmitted information” (to use the authors' terminology) could probably have been imparted in a much slimmer volume. Some of the commentary, however well meant, will be mainly of academic interest to most computer users, since they are stuck with the tools that have been provided for them. I doubt that one of the intended audiences for this book—engineers (unless they are software engineers developing graphics packages)—would bother looking here for inspiration on how to present information. Most engineers are computer users, not software developers, and will not be sitting down to write a CAD or scientific visualization program from scratch. The amount of effort and resources required to do so would be staggering, with very little potential benefit, as compared with using the wealth of high-performance commercial software available. As for the other target audiences—computer scientists and researchers from other disciplines who are interested in pictures—they may find something of interest here, but the amount of information that they can acquire from this book is relatively low. This book does provide some food for thought, but if you are looking for real substance, it does not have much. Many terms and ideas are bandied about that seem better suited to books on artificial intelligence, information theory, art theory, psychology, and philosophy. It is interesting to discuss these ideas under the umbrella notion of “interactive systems,” but much of the discussion is not pertinent to interactive systems as a whole, rather only to how we might interact with computers in a general sense. Pictures do play an important role in human-computer interface, knowledge systems, expert systems, and in so many different areas, but in this book they are overplayed. Pictures can convey a great deal of information, and this important concept is extensively explored; a distinction is drawn between directly imparted information (transmitted) and derived information (transputed). The clear distinction between data and information is not defined, however; indeed, the two terms are confused occasionally. For example, what is “semantic information,” as distinct from “semantic data”__?__ Information is the useful stuff that is derived from data. There is a wealth of data available in the universe, but far less information. The authors ask why there is no universal language for defining pictures, which might then form the basis for a knowledge structure about the pictures. Most pictures are produced as sets of pixels, and a pixel map is an unintelligent knowledge structure about the picture. There is information content within the data (in a Shannonian sense), but so what__?__ This is not particularly useful in determining the real nature of the picture or the model from which it is generated. The picture itself is just a mapping of the real knowledge about something, and this information needs to be encoded in a much more high-level, structured way. The authors use the term “oracle” to refer to information input by a user involving some synergistic relationship between computer and user resulting in a change in the machine's representation of data. The oracle is explored as a method by which the computer can learn from the information that may be transputed to a user, for example, via a graphical animation during a simulation sequence. The idea is based on questions about what we know about what learners see when looking at pictures. The authors argue that the computer should be given this information somehow (by means of a “viewpoint description”), so that it can gain a little extra knowledge about a problem that might benefit others using the system. The approach is to either employ a special editor that allows a user to criticize and document the quality of the result, or to use a graphical interface to point at picture components (preferred). A formal description is generated that embodies the essential information that has been transputed to the user and this, in turn, can somehow be fed back into a knowledge struc<__?__Pub Caret>ture so that the computer model is also cognizant of the information. Pictures play an important role in the communication of information, but computer-generated images should not always be seen as the be all and end all. Like the computer itself, they are only tools meant to assist users in solving problems. Interactive graphics is really nothing more than modeling a problem with a computer program, generating some graphical output from the model, then changing parameters to alter either the picture or its model description and rerunning the model to produce fresh graphical output. The process is an iterative one, leading users toward optimal solutions to their problems in less time than it would take to do the same thing as a series of batch runs, tuning parameters for each new batch depending on the results produced from the previous run. Half a book is not needed to say this.