The eigenvalues of the Dirichlet Laplacian are used to generate three different sets of features for shape recognition and classification in binary images. The generated features are rotation-, translation-, and size-invariant. The features are also shown to be tolerant of noise and boundary deformation. These features are used to classify hand-drawn, synthetic, and natural shapes with correct classification rates ranging from 88.9% to 99.2%. The classification was done using few features (only two features in some cases) and simple feedforward neural networks or minimum Euclidian distance.
This chapter presents a memoir of Jan Bart Le Poole with particular emphasis on Jan Le Poole's definitive and highly original contribution to the design of electron microscopes and to particle optics in general, at a time when the basic theory and technology were in their infancy. The chapter discusses Jan Bart Le Poole's strong motivation and powers of survival under wartime conditions, together with his creative insights and inventive approach. The impressive attempts in Delft in 1969 by Van Zuylen to cancel spherical aberration in the transmission electron microscope (TEM) by Gabor's in-line holography did not succeed in improving the resolution, for reasons that are now quite clear but were not so obvious at the time. It is perhaps significant to recall that Hannes Lichte chose a Philips column for his attempt to achieve aberration-free atomic resolution with computer reconstruction—a tribute to Jan Le Poole's foresight in providing an outstanding column design for this purpose. The only design modification needed was the insertion of an electron biprism in place of one of the selected-area apertures. With no other changes, off-axis holography could be performed at atomic resolution.
Publisher Summary This chapter describes the way in which an annular dark-field (ADF) image is formed in a scanning transmission electron microscope (STEM). ADF imaging refers to the use of particular detector geometry in STEM. A geometrically large annular detector is placed in the optical far field beyond the specimen. The total intensity detected over the whole detector is recorded and displayed as a function of the position of the illuminating probe. Because the detector only receives a signal when the specimen is present, the vacuum appears dark, hence the name, and the heavier the atom, the higher the intensity of the scattering, which leads to atomic number (Z) contrast in the image. The most important feature of ADF imaging is that it can be described as being incoherent that has many advantages at atomic resolution. The chapter explains the way in which the image data may be used to provide atomic-resolution information about the specimen.
Publisher Summary This chapter discusses the role played by tools originating from the field of artificial intelligence (AI). The aim of AI is to stimulate the developments of computer algorithms able to perform the same tasks that are carried out by human intelligence. Some fields of application of AI are automatic problem solving methods for knowledge representation and knowledge engineering, machine vision and pattern recognition, artificial learning, automatic programming, the theory of games, and so forth. Although methods based on the signal theory and the set theory remain the most frequently used methods for the processing of microscope images, methods originating from the framework of AI and pattern recognition seem to produce a growing interest. Among these methods, some of those related to automatic classification and to dimensionality reduction are already being used extensively. The chapter evaluates whether or not tools originating from pattern recognition and AI have diffused within the community of microscopists. It discusses methods available for image processing and analysis in the framework of pattern recognition and AI.
This chapter discusses two approaches: (1) development of the theory of the filter, and (2) applications of the filter. The development of the theory is approached from two perspectives, statistical and logical. The chapter presents the statistical analysis concerning the probability distribution of the filtered data and the associated signal-to-noise ratio (SNR). The filter has not been widely used for noise reduction, and SNR theory of the filter has not been fully described. Logical theory leads to the understanding of the way in which the filter functions to enable its use as a tool based on the morphology of the roots of the filter in one, two, and possibly three dimensions. The logical analysis of the filter is much richer in content and leads to the logical theory of the ranked-order filter. The roots of the filter are a relatively small number of patterns of data satisfying a relational structure and are transparent to the filter. As these roots are defined by a relational structure, the co-joining or conjunction of them is similarly constrained by a relational structure.
The compound optical microscope has been central to countless advances in biological research. The growth of cell and molecular biology in recent years has forced the pace of microscopy research to provide high contrast images of increasingly more specific structures and functions in biological specimens. One technique is fluorescence microscopy, where highly specific chemical probes are used to stain particular components of the sample and imaged by their characteristic emission spectra. The method has advanced rapidly through parallel developments in optical filters and lenses, laser and nonlaser light sources, and highly efficient detectors. Fluorescence microscopy is able to deliver the high contrast needed to distinguish the fluorescence light of stained structures from a background of unwanted nonspecific fluorescence and reflected light features by the use of high-specification interference filters. Normally, a fluorescent molecule will absorb a single photon of the illuminating light and be excited to a highly energetic state. After a short but random period of time, some of this energy is lost as the molecule relaxes to a slightly lower energy before the molecule returns to the original or ground state by emitting a photon of light.
A special class of subsets of binary digital images called "well-composed sets" is defined. The sets of this class have very nice topological properties; for example, the Jordan Curve Theorem holds for them, their Euler characteristic is locally computable, and they have only one connectedness relation, since 4- and 8-connectedness are equivalent. This implies that many basic algorithms used in computer vision become simpler. There are real advantages in applying thinning algorithms to well-composed sets, For example, thinning is an internal operation on these sets and the problems with irreducible "thick" sets disappear. Furthermore, we prove that the skeletons obtained are "one point thick" and we give a formal definition of this concept. We also show that these skeletons have a graph structure and we define what this means.
This article gives an overview of a diverse selection of currently used second-generation image coding techniques. These techniques have been grouped into similar categories in order to allow a direct comparison among the varying methods. An attempt has been made, where possible, to expand upon and clarify the details given by the original authors. The relative merits ans shortcomings of each of the techniques are compared and contrasted.
