Search

The paper presents an adaptive scheme for image-data compression. It is a region-based approach that suitably integrates two different approaches to image coding, vector quantization (VQ) and polynomial approximation (PA). The scheme is adaptive from the point of view of the human observer: the perceptually most significant areas are those near edges or details. In smoothed areas, PA can be used with notable results, but there VQ must be employed to ensure high fidelity. The two techniques exhibit a complementarity in both advantages and drawbacks. PA is not efficient in compressing high-frequency areas, but yields the best results when applied to highly correlated data. VQ is unable to reach high-compression ratios because of its low adaptability, but is quite suitable for compressing uncorrelated data. The means to achieve the integration of the two techniques is a control image containing information about edge and texture locations. In the paper, edge encoding and restoration are also addressed, which are closely related to the proposed hybrid scheme; block prediction is also utilized to further reduce the residual redundancy between VQ blocks. The exploitation of the best features of both approaches results in high compression factors, and in perceivable good quality. In particular, bit rates range from 0.15 to 0.07 bpp. Main applications of this compression scheme are in the areas of very-low bit rate image transmission and image archiving. >

Image processing has frequently to deal with thin structures that hold the information required. In natural images, there is a wide variability of conditions and the borders of objects are not always well defined. Progressive scan, an algorithm derived from fuzzy-connectedness theory, is described here as a specific case of more general /spl chi/-connectedness. Exploiting some analogies to the mechanical world, it allows one to track thin structures and consequently, is suitable for river, motorway or vessel extraction. In the computation of fuzzy connectedness, a widely used approach is based on an adaptive growing mechanism that follows the best paths starting from a reference seed point. The new algorithm uses three or more seed points in order to better drive that approach along the structure of interest. Results of the algorithm are the best path along an object and a connectivity map. The algorithm can deal with heterogeneous images ranging from medical to remote sensed ones.

The system has been developed as a tool to measure in patients the evolution of idiopathic scoliosis during the age of development. The correspondent anatomical segment is treated as a 3D structure, because the interaction of both pathological components and physiological sagittal curvatures determines a well defined location in space for each vertebra involved in the scoliotic curve. The system is based on the acquisition of photographic orthogonal projections, directly from the patient, therefore avoiding X-ray exposure. According to a stack-like scheme, it is possible to obtain all the reconstructed sections, at various height. Various parameters are measured on the sections: in order to discriminate between normal and pathological "scoliotic" sections, the variations of the position of the center of mass and the inertia principal axes appear of importance. Eventually, full 3D capability of the "reconstructed" rib cage, from different points of view, is possible for the clinician.

The present work is focused on design and development of an architecture able to provide a remote segmentation service for various kinds of images or applications. The system exploits the functionalities offered by the already existing desktop application Isocontorni, by extending its capabilities towards client/server environment. In order to achieve this goal the original Isocontorni code has been decomposed in independent modules for processing, storing and representing purposes. Using the created server components a Web application has been developed to demonstrate how to make the fuzzy image segmentation service highly available through Internet. Performances evaluation in terms of elaboration times with different image dimensions have been performed in order to compare the Web-based solution between the pc-based one and to prove the reliability of the proposed system

Summary form only given. Segmentation algorithms have been developed in order to analyze not only intensity images but also some virtual ones, obtained either by histogram-based techniques or by an adaptive technique using the fractal approach. With regard to this last method, the fractal dimension has been used as a parameter that represents the local roughness of a surface. The first operation to be performed is the smoothing of the input images to reduce speckle noise. This has been done with a rank-order operator, which is segmentation-oriented, in that it allows the locally most representative grey value from an ordinated set of grey levels in a processing window to be extracted; therefore, new grey levels cannot be generated. >

The problem of modeling 3-D real objects by using superquadric primitives is addressed. A method for automatic decomposition of complex 3-D shapes is presented. The capabilities of the method are evaluated by applying it to real 2-D and 3-D images of the human brain, which have been obtained from magnetic resonance multislices. The results show the feasibility of the approach and also confirm its capability for quantitatively evaluating the results by means of error-rate computation. >

The application of structured neural networks to the supervised classification of multisensor images is discussed. The purpose is to give a criterion for network architecture definition and to allow the interpretation of the network behavior. The latter result can be used to understand the importance of sensors and related channels to the classification task. The networks' architecture is configured by exploiting the characteristics of a given multisensor classification problem. Then, such networks are trained to solve the problem. Finally, they are transformed into equivalent networks to obtain a simplified representation. >

An advanced vector quantizer for coding true-color images is presented. The starting point is the classical algorithm proposed by Y. Linde, A. Buzo, and A.M. Gray (LBG, 1980). On the basis of the LBG algorithm, a neural net able to furnish a starting codebook quite close to the optimal one is developed. As several vectors are infrequently used, a merge-and-split process has been performed to merge such vectors into adjacent ones, thus allowing free space to be reutilized to redistribute the most numerous partitions. Another aspect to be exploited is the existence of statistical redundancies between neighboring blocks. To sharply reduce the bit rate, an RGB predictor based on some transition matrices is developed. The basic idea of the development of an RGB predictor is provided by the address vector quantization (AVQ) technique. Results obtained by processing standard images are discussed. >

Publisher Summary In the surveillance research field applied to public areas, crowd monitoring is very useful but presents particularly complex problems. Recognizing objects and persons, and tracking their movements in complex real scenes by using a sequence of images are among the most difficult tasks in computer vision. Object and human motion tracking in 3D real scenes can be achieved by means of Kahnan filtering—a suitable mathematical model for describing objects and persons—and a refined dynamic model for tracking them while moving and reciprocally interacting is needed. Such approaches can provide accurate and robust results even in uncontrolled real-life working conditions. A method for tracking only a single moving person is presented. The limiting assumption decays, and more general and more complex situations are considered, as several objects moving and interacting in real scenes are treated. It is addressed to present the two main phases as the basis of object recognition and tracking: the selection of a set of image features characterizing each detected mobile object or group of objects, and consequently allowing the system to distinguish an object with respect to another, and the matching procedure which allows one to recognize a certain object even after various frames in which it disappears completely or partially. Because of its real-time functioning, accuracy, and robustness, this method can be used in real-life surveillance systems.

In this paper, after briefly considering the reasons that prevented the development of hybrid systems for signal processing (SP), we point out the requirements for their future exploitation. The need for a better knowledge of different approaches in the scientific community and the definition of methodologies for designing hybrid systems are highlighted as two key points. Then we suggest that the well-known "task-structure analysis" design technique should be modified to make it suitable for hybrid systems. The proposed modification is based on the main need to choose the roles of different approaches and the mechanisms to integrate them. As an example we describe the design of a hybrid system for two-dimensional (2-D) image recognition; the system is based on the integration of a numerical, a symbolic, and a connectionist approach. We detail the integration of the symbolic and connectionist approaches to the generation of the models of the objects to be recognized. We describe the main problems involved and the solutions adopted. In particularly, we exploit the synergistic aspects of the two approaches in order to overcome the bottleneck of knowledge acquisition. Finally, we report experimental results on two applications to show some advantages of the proposed hybrid system.

A system for crowding evaluation in complex environments is presented. The system acquires and processes data from a set of cameras monitoring an underground scene. The processing structure is modelled as a hierarchical Bayesian network of interacting nodes; each node aims at obtaining the probabilistic value of the number of people, detected within either local areas or the whole station, starting from suitable features extracted from images. Piece-wise linear models allow mapping from the feature value space to the number of people to be performed. The modelling algorithm, based on the Bellman Principle, is discussed. Results obtained after an extended test phase in a station of Genova’s underground are reported.

Evaluation of crowding in complex environments is a problem currently addressed in the context of surveillance systems both to detect potentially dangerous situations (overcrowding) and for statistical purposes related to activity planning.

Two novel approaches to interactive image segmentation are presented. Both methods aim to extract the boundary of a connected object. Accuracy and real-time processing are the main objectives to be accomplished. The speed requirement is met by a) drastically reducing the user interaction during processing and, b) performing pseudo 3D processing (propagation of information between slices). Accuracy is improved through multilevel segmentation Results are reported, using the COVIRA [1] reference data set.

Computer-enhanced analysis of electron micrographs of thin-sectioned rat liver nuclei, combined with three-dimensional reconstruction of the same Feulgen-stained nuclei, points to a unique clustering of chromatin DNA fibres near the nuclear border. Computer-enhanced image analysis has been applied to electron micrographs of the envelopes of the same rat liver nuclei prepared by freeze etching and a few essential geometrical parameters characterizing the pores and their distribution have been determined. During interphase, clusters of nuclear pores, closely paralleling the clustering of membrane-attached chromatin fibres, have been identified on the envelope, the number of these being similar to the number of homologus pairs of metaphase chromosomes. Furthermore, rapid changes induced in chromatin distribution appear to be associated with rapid changes in pore number, but not in the number of pore clusters.

The reliability of radiographical information may be improved by a) standardised radiography (exposure-development); b) transformation of the original image into a densitometric image and equalization with a reference wedge made of Ergal. The linear absorption coefficient of Ergal in terms of photon X energy is almost identical to the average coefficient of the bone mineral matrix; c) processing the densitometric image for additional information. An automatic image analyser controlled by a computer is used. The first osteological results showing the correspondence between the densitometric measurements provided by the analyser and the actual bone mineral content revealed by chemical analysis are presented.

A quantitative methodology for the analysis of pulmonary radiological images by using the fractal approach to texture analysis is proposed. Through the analysis of the fractal distribution, some parameters are defined; the histogram is identified with Gaussian and Rayleigh-like functions, in the case of real pulmonary tissue, simulated pathologies, and their overlapped images. By analyzing the superimposition of pathological tissues upon normal ones, it is generally found that the average fractal dimension exhibits an intermediate value between those of the two isolated tissues. Analogously, the variance of an overlapping structure has a value ranging between those for the two original tissues. Since the fractal dimension seems to be more discriminant than intensity information, segmentation can be applied to the extracted images, thus making it possible to isolate normal from pathological areas in the same radiological image. >

In this paper, the focus is on low-level processing of SAR (Synthetic Aperture Radar) images, and the eventual goal is automatic classification employing various techniques based on Mathematical Morphology (MM). SAR images are characterized by considerable "speckle" noise, which gives rise to serious problems in early processing (filtering, edge detection). In order to overcome these problems, we have used the MM approach, in particular, operators for filtering such images to reduce "speckle" noise and to enhance straight lines, typical for man-made objects. Edges are detected and thinned to obtain as many continuous and closed contours as possible. Edge-based segmentation is then performed, and various features are obtained for each region. Moreover, we use and discuss MM tools also to compute the fractal dimension around each pixel with an adaptive technique. Finally, the resulting information is merged to achieve the correct splitting of an image into significant regions, each described by an appropriate set of features (shape, texture, skeleton, linear edges by the Hough transform, etc.), which are employed in the next classification step. Experimental results have been obtained by analyzing SAR images of a ground area in Algeria; they are shown and discussed in the paper.

This paper describes a framework for detecting regions of special interest, like reflexes and shadows in industrial scenes. A new approach involving Artificial Intelligence techniques is tested; in fact procedural approaches exhibit several drawbacks such as lack of flexibility and of transparency in knowledge representation because knowledge is buried in code. A Knowledge-based system is proposed that is strictly interfaced with several low level image processing tools.

The number of classes included in a set of data and the distribution of these classes in a data structure constitute a great deal of information about the data structure. Having obtained clusters of sample points, one can derive statistical information on each cluster and interpret the qualitative features of the system represented by the data. The clustering approach becomes a very important tool when no a-priori knowledge is available 1 . If the data vectors form well structured clusters, i.e. compact and separate from each other, a simple clustering procedure can yield clear-cut and unique clusters. If the distribution of the pattern vectors is near uniform, the obtained clusters must be validated and analysed according to different classifications. Most practical cases lie between these two situations

In production engineering an important area of research concerns the typology of surfaces machined with different cutting processes in order to define relationships between the generation variables, the surface finish parameters and the functional proper ties1,2,3.

This paper deals with digital radiological image processing, with a special emphasis on quantitative measurement extraction aimed at discriminating different lung pathologies. Texture analysis of these images, not easily performed by the human eye, can be carried out by means of fractal techniques, which allow different textures to be characterized by a single parameter--i.e. the fractal dimension. To extract the fractal dimension of an image, the so-called "blanket" method is employed, which allows the average fractal dimension of an image area to be evaluated. The result of this processing is a new image whose points are assigned the average fractal dimension value of the area they belong to. The behaviors of fractal dimension histograms related to various simulations of lung pathologies are notably different, which allows such pathologies to be effectively characterized and differentiated on a quantitative basis. The radiographic images the "blanket" method was applied to are related to simulations of both pathological and healthy tissues. The behaviors of fractal dimension histograms and the average fractal dimension values allow the characterization of the different pathologies both for single tissues and in case of superimposition of a healthy tissue on a pathological one. The promising simulation results have encouraged us to carry out the present experimental investigation in the field of lung neoplasms.

Two-dimensional gel electrophoresis is a powerful tool for determining the protein content of biological samples. At present, however, the great quantity of data to be extracted and examined does not allow its application on a wide scale; this result will be attained through the automation of the entire process. In this paper we analyze some basic problems involved in the automated analysis of electrophoresis images, propose some new solutions, and discuss the results obtained in experimental situations. In particular, spatial resolution is discussed, utilizing the results of spot-detection and segmentation procedures applied to images of three different resolutions. A spot-detection method based on the zero crossing of the first derivative of spot density is proposed. Spot segmentation is performed by testing the signs of second-order incremental ratios of spot density. Overlapped spots are separated by iteratively estimating the isolated spot densities, under the assumptions of density additivity and spot symmetry. Pattern recognition for feature extraction and cluster analysis for detection of spot gel constellations are proposed. Finally, an optical-flow calculation technique is modified for application to gel matching. The results obtained by means of such methodologies are promising and could be utilized by a knowledge-based system for automatic interpretation of electrophoresis images; this is the objective we propose for future research in this field.

We present a novel approach to the change detection problem based on a coarse-to-fine strategy. The basic idea consists in assigning to an efficient preliminary coarse-level detection the task to filter out the well known possible false changes (e.g., those due to camera noise and small displacements, or to scene illumination changes). This provides the subsequent fine-level detection with reliable supermasks of the true changed areas in the scene. In this way, the fine-level detection can "focus the attention" on limited parts of the frames, thus yielding remarkable advantages in terms of computational efficiency. Here, just a coarse-level detection algorithm based on background subtraction and on the concept of structure is presented, to stress that any pixel-level algorithm can be used afterwards and benefit in terms of robustness as well as of computational efficiency.