Abstract: Magnetic resonance (MR) brain section images are segmented and then synthetically colored to give visual representations of the original data with three approaches: the literal and approximate fuzzy c-means unsupervised clustering algorithms, and a supervised computational neural network. Initial clinical results are presented on normal volunteers and selected patients with brain tumors surrounded by edema. Supervised and unsupervised segmentation techniques provide broadly similar results. Unsupervised fuzzy algorithms were visually observed to show better segmentation when compared with raw image data for volunteer studies. For a more complex segmentation problem with tumor/edema or cerebrospinal fluid boundary, where the tissues have similar MR relaxation behavior, inconsistency in rating among experts was observed, with fuzz-c-means approaches being slightly preferred over feedforward cascade correlation results. Various facets of both approaches, such as supervised versus unsupervised learning, time complexity, and utility for the diagnostic process, are compared. >

Abstract: Image segmentation by mathematical morphology is a methodology based upon the notions of watershed and homotopy modification. This paper aims at introducing this methodology through various examples of segmentation in materials sciences, electron microscopy and scene analysis. First, we define our basic tool, the watershed transform. We show that this transformation can be built by implementing a flooding process on a grey-tone image. This flooding process can be performed by using elementary morphological operations such as geodesic skeleton and reconstruction. Other algorithms are also briefly presented (arrows representation). Then, the use of this transformation for image segmentation purposes is discussed. The application of the watershed transform to gradient images and the problems raised by over-segmentation are emphasized. This leads, in the third part, to the introduction of a general methodology for segmentation, based on the definition of markers and on a transformation called homotopy modification. This complex tool is defined in detail and various types of implementation are given. Many examples of segmentation are presented. These examples are taken from various fields : transmission electron microscopy, SEM, 3D holographic pictures, radiography, non destructive control and so on. The final part of this paper is devoted to the use of the watershed transformation for hierarchical segmentation. This tool is particularly efficient for defining different levels of segmentation starting from a graph representation of the images based on the mosaic image transform. This approach will be explained by means of examples in industrial vision and scene analysis.

Abstract: The authors present an approach for the recognition of multiple 3-D object models from three 3-D scene data. The approach uses two different types of primitives for matching: small surface patches, where differential properties can be reliably computed, and lines corresponding to depth or orientation discontinuities. These are represented by splashes and 3-D curves, respectively. It is shown how both of these primitives can be encoded by a set of super segments, consisting of connected linear segments. These super segments are entered into a table and provide the essential mechanism for fast retrieval and matching. The issues of robustness and stability of the features are addressed in detail. The acquisition of the 3-D models is performed automatically by computing splashes in highly structured areas of the objects and by using boundary and surface edges for the generation of 3-D curves. The authors present results with the current system (3-D object recognition based on super segments) and discuss further extensions. >

Abstract: Human values have been increasingly used as a basis for market segmentation. The list of values (LOV) is one common approach to segmentation: typically, marketers use the top-ranked value to assign consumers to segments. Although it is simple to implement, the top-rank approach to values segmentation conflicts with Rokeach's concept of an ordered value system, in which individual values are organized in the context of an overall hierarchy. This study uses a new measurement model that identifies latent (unobserved) value-system segments derived from a ranking of the LOV items. Higher-order value-system segments reflect the reality that multiple values will affect an individual's behavior. A values map is also constructed, which allows dimensions underlying the value-system segments to be identified. Data from a national survey show that the resulting value-system segments and values map have face validity consistent with the psychological structure of human values recently hypothesized by S. H. Schwartz and W. Bilsky.

Abstract: A pattern-oriented segmentation method for optical character recognition that leads to document structure analysis is presented. As a first example, segmentation of handwritten numerals that touch are treated. Connected pattern components are extracted, and spatial interrelations between components are measured and grouped into meaningful character patterns. Stroke shapes are analyzed and a method of finding the touching positions that separates about 95% of connected numerals correctly is described. Ambiguities are handled by multiple hypotheses and verification by recognition. An extended form of pattern-oriented segmentation, tabular form recognition, is considered. Images of tabular forms are analyzed, and frames in the tabular structure are extracted. By identifying semantic relationships between label frames and data frames, information on the form can be properly recognized. >

Abstract: We present a model of sensory segmentation that is based on the generation and processing of temporal tags in the form of oscillations, as suggested by the Dynamic Link Architecture. The model forms the basis for a natural solution to the sensory segmentation problem. It can deal with multiple segments, can integrate different cues and has the potential for processing hierarchical structures. Temporally tagged segments can easily be utilized in neural systems and form a natural basis for object recognition and learning. The model consists of a "cortical" circuit, an array of units that act as local feature detectors. Units are formulated as neural oscillators. Knowledge relevant to segmentation is encoded by connections. In accord with simple Gestalt laws, our concrete model has intracolumnar connections, between all units with overlapping receptive fields, and intercolumnar connections, between units responding to the same quality in different positions. An inhibitory connection system prevents total correlation and controls the grain of the segmentation. In simulations with synthetic input data we show the performance of the circuit, which produces signal correlation within segments and anticorrelation between segments.

Abstract: Segmentation is a key problems in image processing. In the framework of mathematical morphology the traditional method of segmentation is based on the watershed transform. This method may be analysed as a region growing algorithm, starting from a set of markers for all zones of interest. The author extends it to the segmentation of color images and illustrates it, with paintings.< >

Abstract: Multimodality images obtained from medical imaging systems such as computed tomography (CT), magnetic resonance (MR) imaging, positron emission tomography (PET), and single photon emission computed tomography (SPECT), generally provide complementary characteristic and diagnostic information. Synthesis of these image data sets into a single composite image containing these complementary attributes in accurate registration and congruence would provide truly synergistic information about the object(s) under examination. We have developed a new method which produces such correlation using parametric Chamfer matching. The method is fast, accurate, and reproducible. Surfaces ar initially extracted from two different images to be matched using semi-automatic segmentation techniques. These surfaces are represented as contours with common features to be matched. A distance transformation is performed for one surface image, and a cost function for the matching process is developed using the distance image. The geometric transformation includes three- dimensional translation, rotation, and scaling to accommodate images of different position, orientation, and size. The matching process involves searching this multi-parameter space to find the best fit which minimizes the cost function. The local minima problem is addressed by using a large number of starting points. A pyramid multi-resolution approach is employed to speed up both the distance transformation and the multi-parameter minimization processes. Robustness in noise handling is accomplished using multiple thresholds embedded in the multi- resolution search. The algorithm can register both partially overlapped and fragmented surfaces. Manual intervention is generally not necessary. Preliminary results suggest registration accuracy on the order of the voxel size used in the registration process. Computational time scales with the number of matching elements used, with about five minutes typical for 2563 images using a modern desktop workstation.© (1992) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

Abstract: An improved model of scenes for image analysis purposes, a nested-hierarchical approach which explicitly acknowledges multiple scales of objects or categories of objects, is presented. A multiple-pass, region-based segmentation algorithm improves the segmentation of images from scenes better modeled as a nested hierarchy. A multiple-pass approach allows slow and careful growth of regions while interregion distances are below a global threshold. Past the global threshold, a minimum region size parameter forces development of regions in areas of high local variance. Maximum and viable region size parameters limit the development of undesirably large regions. Application of the segmentation algorithm for forest stand delineation in TM imagery yields regions corresponding to identifiable features in the landscape. The use of a local variance, adaptive-window texture channel in conjunction with spectral bands improves the ability to define regions corresponding to sparsely stocked forest stands which have high internal variance.

Abstract: Presents a fast segmentation scheme for automatic differential counting of white blood cells. The segmentation procedure consists of three phases. First a novel simple algorithm is proposed for localization of white blood cells. The algorithm is based on a priori information about blood smear images. In the second phase the different cell components are separated with automatic thresholding. The thresholds are selected with a simple recursive method derived from maximizing the interclass variance between dark, gray and bright regions based on the method proposed by Otsu (1979). Finally the segmented regions are smoothed by morphological operations. The segmentation scheme works successfully for classification of white blood cells. Some experimental results are also presented. >

Abstract: Marker segments in travel and tourism like elsewhere are defined either predetermined criteria (a priori segmentation) or result from a clustering process (a posteriori segmentation). Each method requires searching for additional descriptive attributes in order to streamline selective market operation. These descriptors should be much easier to measure than those the segments originate from. For a posteriori segments, in particular, it is desirable to predict segment membership from a set of easily determined consumer attributes without having to process a larger number of psychographics. Prediction of segment membership often employs discriminate analysis. Neural network models are also capable of classifying tourists. Having been trained on a set of input variables (descriptors) and output data (segment membership), they may excel discriminate analysis in determining the correct segment affiliation.

Abstract: This paper proposes a methodology that enables an arbitrary 3-D MRI brain image-volume to be automatically segmented and classified into neuro-anatomical components using multiresolution registration and matching with a novel volumetric brain structure model (VBSM). This model contains both raster and geometric data. The raster component comprises the mean MRI volume after a set of individual volumes of normal volunteers have been transformed to a standardized brain-based coordinate space. The geometric data consists of polyhedral objects representing anatomically important structures such as cortical gyri and deep gray matter nuclei. The method consists of iteratively registering the data set to be segmented to the VBSM using deformations based on local image correlation. This segmentation process is performed hierarchically in scale-space. Each step in decreasing levels of scale refines the fit of the previous step and provides input to the next. Results from phantom and real MR data are presented.© (1992) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

Abstract: The analysis of cardiac magnetic resonance (MR) images and X-rays of bone is considered. Each image type is approached using a different form of fractal parameterization. For the MR images, the goal of the study is segmentation, and to this end small regions of the image are assigned a local value of fractal dimension. For the bone X-rays, rather than segmentation, the large-scale structure is parameterized by its fractal dimension. In both cases, the use of fractals leads to the classification of the parameters of interest. When applied to segmentation, this analysis yields boundary discrimination unavailable through previous methods. For the X-rays, texture changes are quantified and correlated with physical changes in the subject. In both cases, the parameterizations are robust with regard to noise present in the images, as well as to variable contrast and brightness. >

Abstract: This paper presents a method for incrementally segmenting images over time using both intensity and motion information. This is done by formulating a model of physically significant image resgions using local constraints on intensity and motion and then finding the optimal segmentation over time using an incremental stochastic minimization technique. The result is a robust and dynamic segmentation of the scene over a sequence of images. The approach has a number of benefits. First, discontinuities are extracted and tracked simultaneously. Second, a segmentation is always available and it improves over time. Finally, by combining motion and intensity, the structural properties of discontinuities can be recovered; that is, discontinuities can be classified as surface markings or actual surface boundaries.

Abstract: The visualization of 3D phenomena and the extraction of quantitative information from magnetic resonance (MR) image data require efficient semiautomated or automated segmentation techniques. The application of multivariate statistical classification to the segmentation of dual-echo volume data of the human head into tissue types (grey matter, white matter and fluid spaces) is studied in this paper. Tests of the radiometric variability of tissue classes within the data volume demonstrate the improvement of the image acquisition technology and the suitability of statistical methods to perform brain tissue segmentation. Supervised classification is successfully applied to a study of 16 MR volume images of the human head, illustrating the robustness of this method in segmenting brain (white and grey matter) and cerebrospinal fluid (CSF). To avoid subjective criteria involved in the supervised approach, ISODATA clustering as well as clustering based on nonparametric probability density estimation were tested. Both methods performed well (success rates 93.8% and 87.5%, respectively), indicating that the classification procedure can be completely automated. The reproducibility and reliability of supervised and unsupervised classfication were studied by comparing results of segmentation performed by five expert operators. Results suggest that the interoperator and intraoperator variations could be reduced using automated clustering techniques. The accuracy of the volume calculations was quantified by applying the MR imaging and segmentation process to a phantom resembling shape and tissue characteristics of brain tissue. The segmented brain objects are displayed using 3D surface rendering.

Abstract: The author describes a new, robust, parallel algorithm for the segmentation of edge data. The segmentation is in the form of straight-lines and circular arcs. No user-supplied thresholds are necessary, and additionally the further grouping of the segments is simplified. The algorithm has been implemented within the FEX software package for use on sequential machines. Using a number of complex edge maps, he compares the results of the application of the algorithm with that developed by West and Rosin (1991) based on a technique suggested by Lowe (1987).< >

Abstract: Individual viewing decisions have a direct impact on the media planning of television advertisers and, consequently, on the revenues of the major television networks. This paper represents an attempt to better understand these decisions. We use Nielsen people meter data to build a perceptual space for programs. That space is then used to develop models explaining viewers' decision to watch television and their choice of programming. The program-choice model is a clusterwise logit model which searches for segments with similar viewing preferences. A segment-level logit model is then used to model the on-off decision. These models can be used by advertisers and advertising agencies to understand the viewing audience better, and thus to help guide their advertising media placement decisions. The models can also help television networks design programs and program schedules that are more attractive to viewers (and thus advertisers).

Abstract: The system of the present invention applies self-organizing and/or supervd learning network methods to the problem of segmentation. The segmenter receives a visual field, implemented as a sliding window and distinguishes occurrences of complete characters from occurrences of parts of neighboring characters. Images of isolated whole characters are true objects and the opposite of true objects are anti-objects, centered on the space between two characters. The window is moved across a line of text producing a sequence of images and the segmentation system distinguishes true objects from anti-objects. Frames classified as anti-objects demarcate character boundaries, and frames classified as true objects represent detected character images. The system of the present invention may be a feedforward adaption using a symmetric triggering network. Inputs to the network are applied directly to the separate associative memories of the network. The associative memories produce a best match pattern output for each part of the input data. The associative memories provide two or more subnetworks which define data subsets, such as objects or anti-objects, according to previously learned examples. Multi-layer perceptron architecture may also be used in the system of the present invention rather than the symmetrically triggered feedforward adaptation with tradeoffs in training time but advantages in speed.

Abstract: We present a model of how objects can be visually discriminated based on the extraction of depth-from-occlusion. Object discrimination re- quires consideration of both the binding problem and the problem of segmentation. We propose that the visual system binds contours and surfaces by identifying "proto-objects"-compact regions bounded by contours. Proto-objects can then be linked into larger structures. The model is simulated by a system of interconnected neural networks. The networks have biologically motivated architectures and utilize a dis- tributed representation of depth. We present simulations that demon- strate three robust psychophysical properties of the system. The net- works are able to stratify multiple occluding objects in a complex scene into separate depth planes. They bind the contours and surfaces of occluded objects (for example, if a tree branch partially occludes the moon, the two "half-moons" are bound into a single object). Finally, the model accounts for human perceptions of illusory contour stimuli.

Abstract: This paper addresses the problem of object tracking in a sequence of monocular images. The use of regions as primitives for tracking enables to directly handle consistent object-level entities. A motion-based segmentation process based on normal flows and first order motion models provide instantaneous measurements. Shape, position and motion of each region present in such segmented images are estimated with a recursive algorithm along the sequence. Occlusion situations can be handled. We have carried out experiments on sequences of real images depicting complex outdoor scenes.

Abstract: (1992). Fear Appeals: Segmentation is the Way to Go. International Journal of Advertising: Vol. 11, No. 4, pp. 355-366.

Abstract: This paper presents a computational model that segments images based on the textural properties of object surfaces. The proposed Coupled-Membrane model applies the weak membrane approach to an image WI(σ,θ, x, y), derived from the power responses of a family of selfsimilar quadrature Gabor wavelets. While segmentation breaks are allowed in x and y only, coupling is introduced to in all 4 dimensions. The resulting spatial and spectral diffusion prevents minor variations in local textures from producing segmentation boundaries. Experiments showed that the model is adequate in segmenting a class of synthetic and natural texture images.

Abstract: Presents the work performed at the CMM as part of the European PROMETHEUS project. The research deals with the road edge detection, lane segmentation and obstacle recognition in a dynamic scene acquired by a monochrome monocular camera. The image processing is based on morphological segmentation tools. The experiments on over a thousand images show that the approach works well on difficult cases such as dense traffic, and roads without land markers. >

Abstract: A primary stage in the analysis of Synthetic Aperture Radar (and other) images, is to produce a segmentation of the image; that is, a division into sub-regions each, hopefully, representing a single object: a field, hedge, wood etc. Producing such a segmentation is made difficult by the high level of speckle inherent in SAR images, and a number of segmentation techniques have been developed. We describe here a methodology for providing objective tests of segmentation algorithms and give representative resulls for one such algorithm.