This is almost verbatim the text of the conference presented in plenary session PLO6 at the E.C.O. 2 symposium in Paris on april 27, 1989.) I would like to devote this talk to a tutorial review of the domain of optical computing as seen from the viewpoint of optical interconnects. We shall start with a short critical assessment of the state of optical computing, showing that optical interconnects are an important part of it. We shall then review the technologies and the families of components available and whose continuing development appears necessary in this context. The question of interconnects in microelectronics will then be examined, clearly supporting the assertion that optics has a part to play there. We shall close with a review of the fundamental advantages that can reasonably be expected from the introduction of optical interconnects in electronic computers. The general ideas will be illustrated whenever possible by recent examples from the litterature or from projects presently under investigation in our Institute in Orsay. To a large extent, this presentation will elaborate on work inspired by the pionneering 1984 article by Goodman, Leonberger, Kung and Athalel.

Circular scanning of images enables simple feature extraction basically insensitive to object orientation and position. Such is a histogram representing concentric luminous intensity of the image. The histogram shows luminous intensity of concentrinc rings having center in object's centroid. However, features extracted from the image in such a way do not uniquely describe an object. This can cause certain difficulties in pattern recognition stage of image analysis. Latin capitals have been chosen as the objects to be recognized. Two approaches have been analysed. In the deterministic approach pattern recognition is based on matching of the features of an object with those from the feature space. Satisfactory matching results in recognition of the particular capital. Small changes in shapes of the capitals to be recognized can show applicability of the described method in practice. The other approach is based on simple form of adaptive recognition. The result of recognition changes the feature space in order to increase reliability of further recognition. Averaging the data from feature space with the features extracted from the recognized image increase the reliability of further recognition. Experimental results were obtained in problem of recognition of the six capitals given in two different fonts and scales. Images of the capitals were obtained by means of binary semiconductor camera. Image processing, feature extraction and pattern recognition were performed on IBM PC AT.

The paper presents the fundamentals of informative approach to imaging and visual perception of images. The principle of information conformity based on non-linear, inertia-less transformation of images in total information flow has been formulated. Extreme transformations in five classes have been discovered, among which contrast, hyperbolic and uniform transformations are emphasized.

Color photographs possess a poor stability towards light, chemicals heat and humidity. As a consequence, the colors of photographs deteriorate with time. Because of the complexity of processes that cause the dyes to fade, it is impossible to restore the images by chemical means. It is therefore attempted to restore faded color films by means of digital image processing.

We propose to combine the properties of color image correlation with the flexibility of synthetic discriminant filters to generate filters capable of solving various color pattern recognition problems. We show that in addition to being color sensitive, the resulting complex composite filters yield an improvement in recognition accuracy with respect to conventional intensity filters.

This paper describes the results of a work in the field of dynamic scene analysis and more specifically dealing with target tracking from infrared imagery. In order to meet the constraints of future real systems, we developped an approach based on the matching of a discrete model of the mobile relying on unsophisticated segmentation features. After having specified a model of regions in the image, based on physical considerations relative to infrared imagery, we describe a cooperative segmentation algorithm using both region growing and points of interest, and the tracking algorithm capable of adjusting to partial occlusions of the target during a given operational scenario. At last, the limitations of our algorithm are discussed.

Many examples of clutter textures in coherent imaging demonstrate K-distributed statistics which result from a gamma-distributed surface cross-section. Such a surface can be completely described in terms of the average two-point statistics (eg the autocorrelation function, ACF) of the process. We compare simulation methods based on such a noise model with a direct model in which we attempt to represent the underlying properties of the surface. A particular example of random woodland clutter is selected in which case this direct model represents quantities such as tree position, size distribution, shape, volume scattering properties and the illumination. We demonstrate that this model predicts processes, such as shadowing, directly, unlike the noise models which combine all factors into the global description. However, further work is needed on this approach before the predicted textures resemble the original SAR texture as closely as those generated by the noise model when the intensity ACF of the texture is reproduced.

To make biological interpretation of Analytical Ion Microscopy (AIM) images possible, computerised superimposition of ion and optical images has become necessary. By these means, appropriate identification of morphological structures is achievable and functional activity of a tissue can be assessed. Although a sequence of ion images of different mass numbers refers to the same field, slight topographic differences can be observed. They often result in image-to-image misregistrations and improper localisation of chemical distributions with respect to other ones. A method is proposed to register ionic images in respect to the optical image of the serial cross section of the same field.

In texture analysis, the extraction of features from a textured region is generally related to a surface element. The aim of this paper is to introduce a new method using a line element instead of a surface element and to use the corresponding texture features extracted for the segmentation of textured images. First we extract the lines of the initial image, either by the means of an edge detection method, or by a segmentation method based on regions, followed by a thinning operation. Next the texture features are extracted from the lines : these features are line primitives such as terminal points, intersections between lines, "middle" points (points which are neither terminal points nor intersection points), critical points (inflexion, high curvature) and also line segments and their associated measures : length, orientation, contrast. The components of the feature vector characteristic of the texture on a [-N,+N]*[-N,+N] neighbourhood is then defined as the number of occurrence of each primitive on this neighbourhood. The last stage in order to obtain the boundaries consists in defining a distance measuring the difference between feature vectors and in detecting the points where the distance between vectors is higher than a threshold.

A real-time VME-bus preprocessor is presented for the extraction of a symbolic scene description from grey-level images. The hardwired pipeline processor consists of two stages. The first stage extracts the thinned greylevel contours using the local edge direction and a novel gradient maximum tracker which does not rely on thresholding. The second stage of the contour vectorizer breaks down the pixel-based contour image into a set of vectors describing the contours in terms of straight-line polygones.

This paper proposes a novel approach for using boundary moments in pattern recognition, in which the complex moments descriptor of the one dimensional complex function representation of the contour is the feature vector. Experimental results presented show that the method has better performance characteristics in comparison with other methods.

Human subjects easily perceive and extensively use shape regularities such as symmetry or periodicity when they are confronted with the task of object description and recognition. A computer vision algorithm is presented which emulates such behaviour in that it similarly makes use of shape redundancies for the concise description and meaningful segmentation of object contours. A shape complexity measure is proposed, based on making such structure explicit, and therefore akin to algorithmic information theory.

A synthesis procedure for superresolution of an incoherent optical system is analyzed and experimentally implemented. The method uses two different illuminations to allow different portions of the spectrum to pass through the system pupil. Postprocessing of both images gives rise to an increase of resolution.

A new method for three-dimensional object space reconstruction from optically or opto-electronically recorded images and extensive experience with its practical implementation is reported. The new method improves the reconstruction accuracy by an order of magnitude when compared with conventional techniques. It is based on a fundamental modification of the Direct Linear Transformation (DLT) technique of Marzan and Karara. In the latter technique, certain orthogonality conditions are violated due to the presence of a redundant DLT parameter. The elimination of this parameter by means of a non-linear constraint, the additional elimination of certain implicit variables from one side of the non-linear DLT formulation, and the simultaneous redefinition of the corresponding objective function to be minimized permits the development of new algorithms whose practical application results in a dramatic improvement in reconstruction accuracy. The highest accuracy obtained was 0.035 %, or one part in 2860 (0.733 mm rms mean error). Such super-high precision in essential if, for instance, high-quality first and second time derivatives are to be computed from time sequences of reconstructed spatial points.

In computer vision, model conceptions are commonly utilized only at the higher levels of processing. In this article it will be shown that a physically oriented model at the lowest level of processing may serve to gain quantitative statements of recognition performance. To find the limitations of recognition and measurement of occluding edges, a model of their optical properties, in connection with a sensor model was developed. By applying this model conception, without the necessity of referring back to heuristics, it is possible to determine in real images the presence, the position, and the direction of occluding edges, even though if only very small image areas (i.e. 3*3 pixels) are evaluated.

By dynamical encoding of the pupil it is possible to increase image resolution when it is limited by aberrations. This technique can be applied when a photodiode array is used, providing digitized data. This encoded detection must be followed by a decoding transform. A gain in signal-to-noise ratio is obtained with regard to aperture limitation. Application to astigmatism which needs a one dimensional encoding only is presented.

Composite images of the inner coma of comet Halley are constructed in superposing the monochromatic charts transmitted by the three-channel spectrometer during the encounter session on March 9, 1986. The instrument covers the 275- to 710- nm range with a spectral reso-lution of 7 nm. The scanned field of view is a rectangle of dimensions 2° x 1,5°. The images are constructed in displaying the signal of a single pixel or the average value of several pixels. In a second step, combinations of operations are used to interpret the images. At a given wavelength, two luminous components are discriminated: the solar scattered continuum and the selective emission of molecular species. The images provide the spatial distributions of dust an of the OH, CN and C2 molecules. The spatial extent is slightly different from a species to another. The distribution of molecular emissions show evidence of gaseous jets which extend to distances longer than 40000 km. The color of dust may be depicted and appears to be blue at the location of jets.

The Fourier Transform Method (FTM) of fringe pattern analysis is one of the most important phase measuring techniques. Its chief advantage is that it requires only one interferogram which is crucial for industrial inspection. The goal of a fully automated phase retrieving technique, however, remains still elusive. Several refinements in 2-D FTM are proposed to achieve fully automatic, high-accuracy analysis of a fringe pattern. The new possibilities of on-line analysis ensuing from the recent advances in microcomputer technology are discussed.

Far-field diffraction patterns of a glass cylinder are spatially sampled with a photodiode array. Three photometric signal processing methods are discussed. The first one is based on the analysis of fringes in the central lobe. It can be used for diameter monitoring, but not for a wide range of diameter measurements. The second method consists of best fitting between the theoretical model of the irradiance distribution in the central fringe and the corresponding experimental data. The accuracy of this method (± 1μm, in the range 10-50μm) is improved by introducing the OTF of the photosensitive area in the calculation. The third method uses a reliable estimation of the intensity at the center and the half-width of the central fringe in order to resolve the inverse problem by Newton's method. This method can be used for real-time measurements but must be improved by statistical approaches.

A non-perspective optical precise dimensional measurement method for some kind of solid work constructed of metallic glossy planes using a series of analog image operating and digital image processing based on a new concept is developed. Dimensions having just rough tolerance and precise tolerance are measured by two different method in a measurement system. The precision between two groung planes determined by the depth measurement on the optical axis of a TV camera is about 2.5x10-3% /1300mm range. By using an analogue data processing apparatus newly developed, the repeatability on measuring data is improved very much.

A new approach to determine motion from multiple images of a sequence is presented. Motion is regarded as orientation in a three-dimensional space with one time and two space coordinates. Fast cascaded filter operations analyze the orientation. The algorithm is analogous to an eigenvalue analysis of the inertia tensor in three dimensions. Four types of orientation are classified: a) constant regions, where no velocity determination is possible; b) edges, where the velocity component perpendicular to the edge is determined; c) corners, where both components of the velocity vector are calculated; d) motion discontinuities, which are used to mark the boundaries between objects moving with different velocities. The accuracy of the new algorithm has been tested with artifi cially generated image sequences with known velocity vector fields. The tests prove that accurate velocity estimates are gained. Even for a low signal to noise ratios, the standard deviation of the displace ment vectors between two consecutive images is only about 0.1 pixel distances.

A novel method is presented for the determination of position and orientation of interconnected human body segments relative to a spatial coordinate system. The development of this new method was prompted by the inadequacy of the techniques currently in use for recorded images. In these techniques, markers are fixed to certain points on the skin of the subject. However, due to skin movement relative to the skeleton and various other factors, the configurational coordinates derived from digitized marker positions may be grossly erroneous with disastrous consequences for the subsequent motion analysis. The new method is based on body-segment shape recognition in the video-image domain. During the recording session, the subject carries special, tight-fitting clothing which permits the unambiguous recognition of segmental shapes and boundaries from the recorded video images. The recognition is performed by means of an edge detection algorithm followed by the computation of the positions and orientations relative to the spatial axes system of all segments of the body model. The new method is implemented on an advanced, special high speed graphic system (Impuls, System 2400) based on transputer chips. The parallel processing capability of this system permits the simultaneous computation of the configurational characteristics for all segments visible in the image. After processing one complete image frame, the video digitizer is instructed to automatically proceed to the next frame, thereby enabling the user to automatically evaluate large amounts of successive frames.

Fringe-carrier method is used in automatic fringe-order numbering of dynamic in-plane moire patterns. In experiment both static carrier and dynamic moire patterns are recorded. The image files corresponding to instants are set up to assign fringe orders automatically. Subtracting the carrier image from the modulated ones, the moire patterns due to the dynamic deformations are restored with fringe-order variation displayed by different grey levels.