Showing papers on "Image segmentation published in 1986"

Extracting Straight Lines

Abstract: This paper presents a new approach to the extraction of straight lines in intensity images. Pixels are grouped into line-support regions of similar gradient orientation, and then the structure of the associated intensity surface is used to determine the location and properties of the edge. The resulting regions and extracted edge parameters form a low-level representation of the intensity variations in the image that can be used for a variety of purposes. The algorithm appears to be more effective than previous techniques for two key reasons: 1) the gradient orientation (rather than gradient magnitude) is used as the initial organizing criterion prior to the extraction of straight lines, and 2) the global context of the intensity variations associated with a straight line is determined prior to any local decisions about participating edge elements.

Detection of Intensity Changes with Subpixel Accuracy Using Laplacian-Gaussian Masks

Abstract: We present a system that takes a gray level image as input, locates edges with subpixel accuracy, and links them into lines. Edges are detected by finding zero-crossings in the convolution of the image with Laplacian-of-Gaussian (LoG) masks. The implementation differs markedly from M.I.T.'s as we decompose our masks exactly into a sum of two separable filters instead of the usual approximation by a difference of two Gaussians (DOG). Subpixel accuracy is obtained through the use of the facet model [1]. We also note that the zero-crossings obtained from the full resolution image using a space constant ? for the Gaussian, and those obtained from the 1/n resolution image with 1/n pixel accuracy and a space constant of ?/n for the Gaussian, are very similar, but the processing times are very different. Finally, these edges are grouped into lines using the technique described in [2].

Analysis of ultrasound image texture via generalized rician statistics

Abstract: Tissue signatures are obtained from the first- and second-order statistics of ultrasonic B-scan texture. Laboratory measurements and early clinical results show that the image may be segmented to discriminate between different normal tissues and to detect abnormal conditions based on a multidimensional feature space. These features describe the intrinsic backscatter properties of the tissues imaged and may be used as the basis of an automatic tissue characterization algorithm.

Low-Level Segmentation of Aerial Images with Fuzzy Clustering

Abstract: A low-level segmentation methodology based upon fuzzy clustering principles is developed. The approach utilizes region growing concepts and a pyramid data structure for the hierarchical analysis of aerial images. It is assumed that measurement vectors corresponding to perceptually homogeneous regions cluster together in the measurement space. The fuzzy c-means (FCM) clustering algorithm is used in the formulation. Utilization of the fuzzy partitioning allows one to derive a correspondence between the cluster membership function values and (the proportions of) the classes constituting a region. Thus cluster membership values can be used to split mixture regions into smaller regions at a higher resolution level. The feasibility of the methodology is evaluated using a three-channel Landsat image. The results show that the FCM clustering can be used in the single-level segmentation; and that cluster membership function values derived using this algorithm can be utilized effectively as indicators of region homogeneity.

A Region-Based Approach to Digital Image Registration with Subpixel Accuracy

Abstract: Automatic registration of images with translational, rotational, and scaling differences is discussed. To register two images from the same scene, first, the images are segmented and closedboundary regions in the images are extracted. Next, centers of gravity of closed-boundary regions are taken as control points and correspondence is established between the control points. Using this correspondence, the original images are then revisited and the segmentation process is refined in such a way that the obtained corresponding regions become optimally similar. This enables determination of centers of gravity of the regions up to subpixel accuracy. Finally, registration parameters are determined by the least squares error criterion.

An Image Understanding System Using Attributed Symbolic Representation and Inexact Graph-Matching

Abstract: This paper presents a powerful image understanding system that utilizes a semantic-syntactic (or attributed-synibolic) representation scheme in the form of attributed relational graphs (ARG's) for comprehending the global information contents of images. Nodes in the ARG represent the global image features, while the relations between those features are represented by attributed branches between their corresponding nodes. The extraction of ARG representation from images is achieved by a multilayer graph transducer scheme. This scheme is basically a rule-based system that uses a combination of model-driven and data-driven concepts in performing a hierarchical symbolic mapping of the image information content from the spatial-domain representation into a global representation. Further analysis and inter-pretation of the imagery data is performed on the extracted ARG representation. A distance measure between images is defined in terms of the distance between their respective ARG representations. The distance between two ARG's and the inexact matching of their respective components are calculated by an efficient dynamic programming technique. The system handles noise, distortion, and ambiguity in real-world images by two means, namely, through modeling and embedding them into the transducer's mapping rules, as well as through the appropriate cost of error-transformation for the inexact matching of the ARG image representation. Two illustrative experiments are presented to demonstrate some capabilities of the proposed system. Experiment I deals with locating objects in multiobject scenes, while Experiment II is concerned with target detection in SAR images.

ANGY: A Rule-Based Expert System for Automatic Segmentation of Coronary Vessels From Digital Subtracted Angiograms

Abstract: This paper details the design and implementation of ANGY, a rule-based expert system in the domain of medical image processing. Given a subtracted digital angiogram of the chest, ANGY identifies and isolates the coronary vessels, while ignoring any nonvessel structures which may have arisen from noise, variations in background contrast, imperfect subtraction, and irrelevent anatomical detail. The overall system is modularized into three stages: the preprocessing stage and the two stages embodied in the expert itself. In the preprocessing stage, low-level image processing routines written in C are used to create a segmented representation of the input image. These routines are applied sequentially. The expert system is rule-based and is written in OPS5 and LISP. It is separated into two stages: The low-level image processing stage embodies a domain-independent knowledge of segmentation, grouping, and shape analysis. Working with both edges and regions, it determines such relations as parallel and adjacent and attempts to refine the segmentation begun by the preprocessing. The high-level medical stage embodies a domain-dependent knowledge of cardiac anatomy and physiology. Applying this knowledge to the objects and relations determined in the preceding two stages, it identifies those objects which are vessels and eliminates all others.

Graph theory for image analysis: an approach based on the shortest spanning tree

Abstract: The paper describes methods of image segmentation and edge detection based on graph-theoretic representations of images. The image is mapped onto a weighted graph and a spanning tree of this graph is used to describe regions or edges in the image. Edge detection is shown to be a dual problem to segmentation. A number of methods are developed, each providing a different segmentation or edge detection technique. The simplest of these uses the shortest spanning tree (SST), a notion that forms the basis of the other improved methods. These further methods make use of global pictorial information, removing many of the problems of the SST segmentation in its simple form and of other pixel linking algorithms. An important feature in all of the proposed methods is that regions may be described in a hierarchical way.

Segmentation of textured images using Gibbs random fields

Abstract: A new algorithm for the segmentation of textured images is developed by making use of Gibbs random fields. A hierarchical stochastic model is employed to represent textured images. At the higher level, the region formation process, describing different areas of the image, is modeled as a Gibbs random field, or equivalently as a Markov random field. At the lower level, the textures in different regions of the image are modeled also as Gibbs random fields. Based on this hierarchical model, the segmentation algorithm being proposed seeks to obtain the maximum a posteriori estimate of the region process using the textured image data. The maximization is carried out recursively by making use of a dynamic programming formulation. Computational concerns, however, necessitate the implementation of a suboptimal version of the algorithm that tries to maximize a pseudolikelihood over strips of the image. This is a non-trivial extension of a maximum a posteriori segmentation algorithm for noisy images modeled by Gibbs random fields [1]. The segmentation algorithm is applied on several textured images composed of 2, 3 region (texture) types and 2 or 4 level textures, with remarkable success. Numerous examples on the application of the segmentation algorithm are presented for textured images with region processes and textures generated according to a particular Gibbs distribution.

Segmentation of a Thematic Mapper Image Using the Fuzzy c-Means Clusterng Algorthm

Abstract: In this paper, a segmentation procedure that utilizes a clustering algorithm based upon fuzzy set theory is developed. The procedure operates in a nonparametric unsupervised mode. The feasibility of the methodology is demonstrated by segmenting a six-band Landsat-4 digital image with 324 scan lines and 392 pixels per scan line. For this image, 100-percent ground cover information is available for estimating the quality of segmentation. About 80 percent of the imaged area contains corn and soybean fields near the peak of their growing season. The remaining 20 percent of the image contains 12 different types of ground cover classes that appear in regions of diffferent sizes and shapes. The segmentation method uses the fuzzy c-means algorithm in two stages. The large number of clusters resulting from this segmentation process are then merged by use of a similarity measure on the cluster centers. Results are presented to show that this two-stage process leads to separation of corn and soybean, and of several minor classes that would otherwise be overwhelmed in any practical one-stage clustering.

Progress in robot road-following

Abstract: We report progress in visual road following by autonomous robot vehicles. We present results and work in progress in the areas of system architecture, image rectification and camera calibration, oriented edge tracking, color classification and road-region segmentation, extracting geometric structure, and the use of a map. In test runs of an outdoor robot vehicle, the Terregator, under control of the Warp computer, we have demonstrated continuous motion vision-guided road-following at speeds up to 1.08 km/hour with image processing and steering servo loop times of 3 sec.

Statistical model-based algorithms for image analysis

Abstract: In this paper, two-dimensional stochastic linear models are used in developing algorithms for image analysis such as classification, segmentation, and object detection in images characterized by textured backgrounds. These models generate two-dimensional random processes as outputs to which statistical inference procedures can naturally be applied. A common thread throughout our algorithms is the interpretation of the inference procedures in terms of linear prediction residuals. This interpretation leads to statistical tests more insightful than the original tests and makes the procedures computationally tractable. This paper also examines a computational structure tailored to one of the algorithms. In particular, we describe a processor based on systolic arrays that realizes the object detection algorithm developed in the paper.

Models of Segmentation

Abstract: The formation of segmented structures is a very important step during development of higher organisms. With the formation of somites in vertebrates or the segments in insects the primary anteroposterior pattern of the organisms is laid down. Segmentation is the result of the superposition of two pattern formation processes. One generates a periodic pattern, i.e. a repetition of homologous structures. It consists in vertebrates of somites and somitic clefts and in insects of segments and segment borders. Superimposed on this periodic pattern is a sequential pattern which makes the repetitive subunits different from each other. In recent years, we have proposed molecularly feasible models which are able to generate periodic and sequential structures precisely superimposed on each other (Meinhardt, 1982a,b). For insect development more detailed experimental and genetic data are available. For that reason the model of insect segmentation is more advanced. At the beginning of this paper a short overview of the model proposed for insect segmentation will be provided. I will show that, with some adaptations, the basic mechanism is also able to account for somite formation.

Image Structure Representation and Processing: A Discussion of Some Segmentation Methods in Cytology

Abstract: Image processing methods (segmentation) are presented in connection with a modeling of image structure. An image is represented as a set of primitives, characterized by their type, abstraction level, and a list of attributes. Entities (regions for example) are then described as a subset of primitives obeying particular rules. Image segmentation methods are discussed, according to the associated image modeling level. Their potential efficacity is compared, when applied to cytologic image analysis.

A new thresholding method for volume determination by SPECT.

Abstract: The quantification of organ volumes from SPECT images suffers from two major problems: image segmentation and imperfect system transfer function. Image segmentation defines the borders of an organ and allows volume measurements by counting the voxels inside this contour in all slices containing parts of this organ. A review of the literature, showed that several investigators use a fixed threshold (FT) to determine the organ pixels. It is our aim to demonstrate that the threshold has to be adapted to every single case because its value is dependent upon several factors, such as size and contrast. Therefore a threshold selection algorithm, based on the gray level histogram (GLH), is evaluated. It is nearly impossible to calculate and eliminate errors induced by the complex system response function. A correction method based on linear regression is proposed. By minimizing the relative error (σ), a linear correlation (Y=AX+B) between the true volume (Y) and the measured volume (X) is established for three fixed thresholds (30%, 40%, 50%) and for the GLH method. The methods are evaluated on a series of nineteen phantoms with a volume range between 9.8 and 202.5 ml. The relative error is minimal for the GLH method. The whole procedure is semi-automated and virtually operator independent.

A Pyramid-Based Approach to Segmentation Applied to Region Matching

Abstract: In this paper, we attempt to place segmentation schemes utilizing the pyramid architecture on a firm footing. We show that there are some images which cannot be segmented in principle. An efficient segmentation scheme is also developed using pyramid relinking. This scheme will normally have a time complexity which is a sublinear function of the image diameter, which compares favorably to other schemes. The efficacy of our approach to segmentation using pyramid schemes is demonstrated in the context of region matching. The global features we use are compared to those used in previous approaches and this comparison will indicate that our approach is more robust than the standard moment-based techniques.

Vision-Based Guidance of an Agricultural Tractor

Abstract: The ordered structure of agricultural row crops can provide useful guidance information for tractor control. A description of research for coupling a machine vision system and a solid state camera to derive vehicle guidance parameters for a tractor is presented. Image segmentation is enhanced by optical filtering and controlling light intensity to the image sensor. An analysis of camera location and steering errors that can be determined from the row crops is determined by simulating the geometric relationships between the crop canopy and the image plane.

Markov random fields for image modelling and analysis

Abstract: This chapter deals with the problem of image modelling through the use of 2D Markov Random Field (MRF). The MRF’s are parametric models with a noncausal structure where the various dependencies over the plane is described in all directions. We first show how the MRF’s are used as texture image models, region geometry models, as well as edge models.Then we show how they have been successfully used for image classification, surface inspection, image restoration, and image segmentation.

An iterative region-growing process for cell image segmentation based on local color similarity and global shape criteria.

Abstract: An image segmentation process was derived from an image model that assumed that cell images represent objects having characteristic relationships, limited shape properties and definite local color features. These assumptions allowed the design of a region-growing process in which the color features were used to iteratively aggregate image points in alternation with a test of the convexity of the aggregate obtained. The combination of both local and global criteria allowed the self-adaptation of the algorithm to segmentation difficulties and led to a self-assessment of the adequacy of the final segmentation result. The quality of the segmentation was evaluated by visual control of the match between cell images and the corresponding segmentation masks proposed by the algorithm. A comparison between this region-growing process and the conventional gray-level thresholding is illustrated. A field test involving 700 bone marrow cells, randomly selected from May-Grunwald-Giemsa-stained smears, allowed the evaluation of the efficiency, effectiveness and confidence of the algorithm: 96% of the cells were evaluated as correctly segmented by the algorithm's self-assessment of adequacy, with a 98% confidence. The principles of the other major segmentation algorithms are also reviewed.

Surfaces in early range image understanding

Abstract: Perception of surfaces plays a fundamental role in three-dimensional object recognition and image understanding. A range image explicitly represents the surfaces of objects in a given field of view as an array of depth values. Previous research in range image understanding has limited itself to extensions of edge-based intensity image analysis or to interpretations in terms of polyhedra, generalized cylinders, quadric primitives, or convex objects. Computer vision research has demonstrated the advantages of data-driven early processing of image data. If early processing algorithms are not committed to interpretation in terms of restrictive, domain-specific, high-level models, the same algorithms may be incorporated in different applications without substantial effort. A general approach has been developed for processing range images to obtain a high-quality, rich (information-preserving), accurate, intermediate-level description consisting of graph surface primitives, the associated segmented regions, and their bounding edges. Only general knowledge about surfaces is used to compute a complete image segmentation; no object level information is involved. The early range image understanding algorithm consists primarily of a differential-geometric, visible-invariant pixel labeling method based on the sign of mean and Gaussian curvatures and an iterative region-growing method based on variable-order surface-fitting of the original image data. The high-level control logic of the current implementation is sequential, but all low-level image processes can be executed on parallel architectures. This surface-based image analysis approach has successfully segmented a wide variety of real and synthetic range images and is also shown to have significant potential for intensity image analysis. It is interesting to note that the surface and edge description algorithms use the same basic "sign-of-curvature" paradigm in different dimensions.

Recursive region splitting at hierarchical scope views

Abstract: There are several limitations of the recursive region splitting algorithm for image segmentation. The recursive region splitting at hierarchical scope view is a new algorithm to ease some of the difficulties. A quad tree structure is used to store the split results of the scope views at different levels. The segmentation will proceed to small scope views only if the result at that level is not satisfactory according to a certain criterion. Experimental results of applying this new algorithm to aerial photographs shows improved sensitivity to detect smaller objects. A boundary check procedure is implemented in this algorithm to remove boundary discontinuity along the scope view border. The segmentation results and processing time of four sets of aerial photographs are also discussed here.

Some pyramid techniques for image segmentation

Abstract: This paper describes a collection of multiresolution, or “pyramid”, techniques for rapidly extracting global structures (features, regions, patterns) from an image. If implemented in parallel on suitable cellular pyramid hardware, these techniques require processing times on the order of the logarithm of the image diameter.

A unified method for segmentation and edge detection using graph theory

Abstract: Methods of image segmentation and edge detection based on graph theoretic representations of images are described. The image is mapped onto a weighted graph and, from this graph, spanning trees are used to describe regions and edges in the image. Edge detection is shown to be a dual problem to segmentation. Two methods are developed for both segmentation and edge detection, providing four related techniques. The simpler method uses the Shortest Spanning Tree (SST) to partition the graph and to form a segmentation or edge detection. The second method applies the first method recursively to incorporate global pictorial information into the graph, removing many problems of the simpler method and of other pixel-linking algorithms. An important property of the segmentation and edge detection methods is that image features may be described in a hierarchical way.

Use Of Texture Operators In Segmentation

Abstract: Computer vision systems applicable for the analysis of complex high resolution aerial images require reliable and robust operators for extracting information from the images. These operators should be able to interrogate the image data and derive meaningful information about the presence of various objects appearing in the scene. In this paper we describe the development of operators that are derived from a texture analysis methodology for performing segmentation of high resolution imagery. The utility of these operators for characterizing various perceptually meaningful properties is demonstrated by performing experimental analysis of urban scenes.

Edge detection using the linear model

Abstract: An edge detector based on the linear model is developed which utilizes the generalized likelihood ratio for statistical hypothesis testing. The detector is invariant to multiplicative changes in the gray-scale values of the image. Hence, thresholding based histogram segmentation is not required. The performance of this detector is analytically and experimentally compared to that of a gradient operator (Sobel) and is shown to have only a slightly poorer detection rate for a given false alarm rate.

Feature-based image segmentation in human vision.

Abstract: Experiments are described which indicate that the integration of high-precision shape information along a bright line is blocked by the presence of certain image features. All the features involved have three properties: (1) they are points where contours are not smooth (i.e. not twice differentiable) within the limits set by the finite space constants of visual processes; (2) they are all points that are emphasized in the responses of certain classes of circularly symmetric bandpass spatial filter; and (3) they are all significant for three-dimensional shape analysis. The results are interpreted as implying an inflexible segmentation of the contour image before detailed shape analysis.

Rule-Based Interpreting Of Aerial Photographs Using The Lockheed Expert System

Abstract: Human photointerpreters use expert knowledge and contextual information to help them analyze a scene. We have experimented with the Lockheed Expert System (LES) to see if contextual information can be useful in interpreting aerial photographs. First, the gray-scale image is segmented into uniform or slowly varying intensity regions or contiguous textured regions using an edge-based segmentation technique. Next, the system computes a set of attributes for each region. Some of these attributes are based on local properties of that region only (e.g., area, average intensity, texture strength, etc.); others are based on contextual or global information (e.g., adjacent regions and nearby regions). Finally, LES is given the task of classifying all the regions using the attribute values. It utilizes multiple goals and multiple rule sets to determine the best classification; regions that do not satisfy any of the rules are left unclassified. The authors obtained the rules by an introspection technique after studying many aerial photographs. Unlike programs that use only statistics in the region under consideration, LES can use contextual information such as the fact that cars are likely to be adjacent to roads, which significantly improves its performance on regions that are difficult to classify.

Low Level Segmentation of Noisy Imagery

Abstract: : A possible approach to image segmentation is first to perform a low- level segmentation. This then allows an original image to be described in terms of a set of simple regions or primitives. Objects in the image may be subsequently recognized by matching these primitives to patterns of primitives in a data base. It is found that current techniques for low-level image segmentation fail when applied to high noise images. An algorithm is presented which overcomes the problems associated with high noise and succeeds in generating low-level segmentations of noisy imagery. The algorithm is shown also to work on low noise data.