Multiresolution image processing and analysis (Q792774)

The book is a collection of papers presented at a workshop held in Leesburg, Virginia in July 1982. Methods of multiresolution image processing and analysis are based on a hierarchical data structure that can be obtained from the original image and are called a ''pyramid''. Prof. Rosenfeld, in the leading article of the book, characterizes the pyramid in the following way: ''The simplest type of image pyramid is constructed by repeatedly averaging the image intensities in nonoverlapping 2-by-2 blocks of pixels. Given an input image of size \(2^ n\) by \(2^ n\), applying this process yields a reduced image of size \(2^{n-1}\) by \(2^{n-1}\); and so on. If we imagine images stacked on top of one another, they constitute an exponentially tapering 'pyramid' of images.'' The book is divided into seven parts. In the introductory part, ''Image pyramids and their uses'', an overview of multiresolution data structures and techniques to handle these structures are given. Two basic types of pyramids are defined, the Gaussian Pyramid (multiresolution low-pass filter) and the Laplacian Pyramid (multiresolution bandpass filter) along with other computational tools, methods of image representation, texture analysis as well as the methods of motion analysis of images. The part ''Architectures and systems'' deals with the hierarchical organization of the small processors for bottom-up image analysis, automated generation of recognition strategies, and practical aspects of developing a multiresolution image processing computer. In the part III, ''Modelling, processing and segmentation'', the time series models for representing multiresolution images, node linking strategies in pyramids for image segmentation, multilevel image reconstruction and SIMD cellular array pyramid machines are discussed. The problems of feature and shape analysis, region representation and surface analysis are covered in parts IV and V. Analysis of time-varying images and dynamic scene analysis is introduced in part VI of the book. The last part is devoted to practical applications of the discussed principles, multiresolution microscopy and two-resolution detection of lung tumors in chest radiographs are shown. The methods introduced in the book are based on a new general approach having significant advantages over the conventional, that is non- multiresolution, methods. The individual papers offer the reader an excellent overview of the novel approach playing an ever more important role in image processing and analysis.