This book is concerned with the aspects of real-time, parallel computing which are specific to the analysis of digitized images including both the symbolic and semantic data derived from such images. The subjects covered encompass processing, storing, and transmitting images and image data. A variety of techniques and algorithms for the analysis and manipulation of images are explored both theoretically and in terms of implementation in hardware and software. The book is organized into four topic areas: (1) theo retical development, (2) languages for image processing, (3) new computer techniques, and (4) implementation in special purpose real-time digital systems. Computer utilization, methodology, and design for image analy sis presents special and unusual problems. One author (Nagao)* points out that, "Human perception of a scene is very complex. It has not been made clear how perception functions, what one sees in a picture, and how one understands the whole picture. It is almost certain that one carries out a very quick trial-and-error process, starting from the detection of gross prominent features and then analyzing details, using one's knowledge of the world. " Another author (Duff) makes the observation that, "It is therefore more difficult to write computer programs which deal with images than those which deal with numbers, human thinking about arithmetic being a largely conscious activity."

In order to rea1ize real-time medica1 imaging systems, such as are used for computed tomography, automated miscroscopy, dynamic radioisotope imaging, etc., special techno1ogy is required. The high-speed image sour ce must be successfu11y married with the u1tra- high-speed computer. Usua11y the ordinary genera1-purpose computer is found to be inadequate to the image generation and/or image pro- cessing task. The ordinary computer executes instructions at be- tween 1 and 10 million per second. Speed has improved by only about a factor of 10 during the past 20 years. In contrast a typical com- puter used in recognizing blood cell images at 10,000 per hour must execute instructions at between 1 billion and 10 billion per second. Simi1ar execution rates are required to construct a computed tomogra- phy image in real-time (1 to 10 seconds). For the reasons given above, engineering development in image generation and processing in the field of biomedicine has become a discipline unto itself; a discipline wherein the computer engineer is driven to design extremely high-speed machines that far surpass the ordinary computer and the x-ray, radioisotope, or microscope scanner designer must also produce equipment whose specifications extend far beyond the state-of-the-art.