Showing papers on "Segmentation-based object categorization published in 1991"
TL;DR: The two-dimensional Gabor filters possess strong optimality properties for this task, and local spatial frequency estimation approaches are suggested that use the responses as constraints in estimating the locally emergent texture frequencies.
Abstract: A model for texture analysis and segmentation using multiple oriented channel filters is analyzed in the general framework. Several different arguments are applied leading to the conclusion that the two-dimensional Gabor filters possess strong optimality properties for this task. Properties of the multiple-channel segmentation approach are analyzed. In particular, perturbations of textures from an ideal model are found to have important effects on the segmentation that can usually be ameliorated by simply preceding the segmentation process by a logarithmic operation and using a low-pass postfilter prior to making region assignments. The difficult problems of space-variant textures and multiple component textures are also considered. Local spatial frequency estimation approaches are suggested that use the responses as constraints in estimating the locally emergent texture frequencies. Complex texture aggregates containing multiple shared frequency components can be analyzed if the textures are distinct and few in number. >
187 citations
07 Oct 1991
TL;DR: In this article, the authors address the problem of motion segmentation using the singular value decomposition of a feature track matrix and show that, under general assumptions, the number of numerically nonzero singular values can be used to determine the count of motions.
Abstract: The authors address the problem of motion segmentation using the singular value decomposition of a feature track matrix. It is shown that, under general assumptions, the number of numerically nonzero singular values can be used to determine the number of motions. Furthermore, motions can be separated using the right singular vectors associated with the nonzero singular values. A relationship is derived between a good segmentation, the number of nonzero singular values in the input and the sum of the number of nonzero singular values in the segments. The approach is demonstrated on real and synthetic examples. The paper ends with a critical analysis of the approach. >
182 citations
TL;DR: In this paper, a method for segmenting video scenes hierarchically into several differently moving objects and subobjects is presented, where both contour and texture information from the single images and information from successive images are used to split up a scene into various objects.
Abstract: This contribution presents a method for segmenting video scenes hierarchically into several differently moving objects and subobjects. To this end, both contour and texture information from the single images and information from successive images is used to split up a scene into various objects. Furthermore, each of these objects is characterized by a transform h ( x,T ) with a parameter vector T which implicitely describes the surface shape and the three-dimensional motion of the objects in the scene. In order to estimate T of these transforms, an efficient algorithm is introduced. Thus, we obtain an object-oriented segmentation and a prediction of the image contents from one image to the next, which can be used in low bit-rate image coding.
177 citations
TL;DR: An adaptive split-and-merge image segmentation algorithm based on characteristic features and a hypothesis model is proposed and one of the key processes, the determination of region homogeneity, is treated as a sequence of decision problems in terms of predicates in the hypothesis model.
128 citations
TL;DR: The image segmentation problem is solved by extracting kernel information from the input image to provide an initial interpretation of the image and by using a knowledge-based hierarchical classifier to discriminate between major land-cover types in the study area.
Abstract: A knowledge-based approach for Landsat image segmentation is proposed. The image segmentation problem is solved by extracting kernel information from the input image to provide an initial interpretation of the image and by using a knowledge-based hierarchical classifier to discriminate between major land-cover types in the study area. The proposed method is designed in such a way that a Landsat image can be segmented and interpreted without any prior image-dependent information. The general spectral land-cover knowledge is constructed from the training land-cover data, and the road information of an image is obtained through a road-detection program. >
123 citations
TL;DR: A 3-D segmentation algorithm is presented, based on a split, merge and group approach, that uses a mixed (oct/quad)tree implementation and a number of homogeneity criteria is discussed and evaluated.
97 citations
TL;DR: A method for automatic segmentation of speech into phones is described, where the incoming utterance is split up into more or less stationary parts, and these stationary parts are labelled as phones using the phonetic transcription of the utterance.
Abstract: A method for automatic segmentation of speech into phones is described. The incoming utterance is split up into more or less stationary parts, and these stationary parts are labelled as phones using the phonetic transcription of the utterance. An implicit segmentation algorithm splits up the utterance into segments on the basis of the degree of similarity between the frequency spectra of neighboring frames. An explicit algorithm does the same, but on the basis of the degree of similarity between the frequency spectra of the frames in the utterance and reference spectra. A combination algorithm compares the two segmentation results and produces the final segmentation. Automatically determined phone boundaries are compared with manually determined ones. The result of a perception test is described. >
97 citations
03 Jun 1991
TL;DR: A split-and-merge algorithm is proposed for the segmentation of the digitized surface of a range image into planar regions, which allows a better adaptation of the range image segmentation to the surface boundaries.
Abstract: A split-and-merge algorithm is proposed for the segmentation of the digitized surface of a range image into planar regions. The geometric data structure used is a triangular tessellation of image domain. This data structure, combined with an adaptive surface approximation technique, allows a better adaptation of the range image segmentation to the surface boundaries. It also provides an efficient neighborhood referencing mechanism, thus resulting in a fast algorithm. >
52 citations
09 Apr 1991
TL;DR: A novel method for outdoor natural color image segmentation for road following is presented and it can be seen that the hue, saturation, and intensity (HSI) system should be used rather than the red, green, blue (RGB) system.
Abstract: A novel method for outdoor natural color image segmentation for road following is presented. From the results of experiments it can be seen that the hue, saturation, and intensity (HSI) system should be used rather than the red, green, blue (RGB) system, and segmentation can be done in S-I space with good performance. An automatic adaptive threshold selection method is developed. The preliminary results indicate the effectiveness and efficiency of this method. >
43 citations
11 Jun 1991
TL;DR: A novel texture segmentation technique for both supervised and unsupervised segmentation is presented, which can reach the global maxima of the posteriori distribution even if the textures are modeled by an MRF model.
Abstract: A novel texture segmentation technique for both supervised and unsupervised segmentation is presented. The textured images under study are modeled by a proposed hierarchical Markov random field (MRF) model. This model is formed by combining the binomial model for textures and the multilevel logistic model for region distributions. The supervised segmentation is achieved by a novel algorithm which can reach the global maxima of the posteriori distribution even if the textures are modeled by an MRF model. For unsupervised segmentation, a novel parameter estimation scheme is proposed for estimating the model parameters directly from a given image. The proposed technique is verified by a variety of textured images, such as synthesized textures, natural textures, and aerial images, in both the supervised and unsupervised segmentation cases. >
TL;DR: This paper presents a parallel 3D image segmentation algorithm which, through the use of α- partitioning and volume filtering , segments 3D images such that the greylevel variation within each volume can be described by a regression model.
Abstract: The development of techniques for interpreting the structure of three-dimensional images, f ( x , y , z ), is useful in many applications. A key initial stage in the signal to symbol conversion process, essential for the interpretation of the data, is three-dimensional image segmentation involving the processes of partitioning and identification . Most segmentation and grouping research in computer vision has addressed partitioning of 2D images, f ( x , y ). In this paper, we present a parallel 3D image segmentation algorithm which, through the use of α- partitioning and volume filtering , segments 3D images such that the greylevel variation within each volume can be described by a regression model . Experimental results demonstrate the effectiveness of this algorithm on several real-World 3D images.
TL;DR: The stereo image pair of a speaking person in front of a stationary background taken by two CCD-cameras is used as an input scene for a new segmentation algorithm based on the phase correlation technique which provides a disparity vector field.
Abstract: The stereo image pair of a speaking person in front of a stationary background taken by two CCD-cameras is used as an input scene for a new segmentation algorithm. The algorithm is based on the phase correlation technique which provides a disparity vector field. A brightness adjustment procedure is performed to provide stereo image pairs suited for the segmentation procedure. Then, a coarse segmentation into background and speaking person is made to achieve a reliable segmentation result. Finally, finer segmentation using a coarse-to-fine control strategy is performed only at the object boundaries. An application is demonstrated by applying a lowpass filter selectively to the background of input sequences for low bit rate image coding algorithms. >
31 Oct 1991
TL;DR: An anatomical knowledge-based system for image analysis that interprets CT/MR (computed tomography/magnetic resonance) images of the human chest cavity is reported, using a priori knowledge in the form of masks to guide the segmentation process.
Abstract: An anatomical knowledge-based system for image analysis that interprets CT/MR (computed tomography/magnetic resonance) images of the human chest cavity is reported. The approach utilizes a low-level image analysis system with the ability to analyze the data in bottom-up (or data-driven) and top-down (or model-driven) modes to improve the high-level recognition process. Several image segmentation algorithms, including K-means clustering, pyramid-based region extraction, and rule-based merging, are used for obtaining the segmented regions. To obtain a reasonable number of well-segmented regions that have a good correlation with the anatomy, a priori knowledge in the form of masks is used to guide the segmentation process. Segmentation of the brain is also considered. >
31 Oct 1991
TL;DR: F'altern recognition techniques were used for segmentation of MR images; the classification was based on multi-spectral image intensities and colorcoded anatomically mapped images.
Abstract: F'altern recognition techniques were used for segmentation of MR images; the classification was based on multi-spectral image intensities. Segmentation of image data into tissue types was done using two supervised artificial neural network algorithms, back-propagation and cascade correlation, and an unsupervised clustering algorithm, fuzzy c-means. Input data consisted of TI-weighted, T2-weighted. and spin-density-weighted images. The segmentation was based on the pixel intensities in each image. With the supervised algorithms, cluster analysis was also run using interslice and interpatient training data. Tissue classification was presented as colorcoded anatomically mapped images.
08 Apr 1991
TL;DR: A new segmentation-based image coding method that adaptability of the partition is achieved by an optimized 2-dimensional piecewise constant approximation of the image, made computationally feasible by a novel preprocessing technique.
Abstract: A new segmentation-based image coding method is proposed. The encoder recursively partitions an image into convex n-gons, 3 >
IBM1
TL;DR: The complete system, which consists of a feature extraction module and the connectionist classifier module, has been designed and implemented in digital VLSl; system architectural aspects as well as the procedure of adaptation of the system to different segmentation problems are discussed.
Abstract: An image segmentation system which uses a connectionist classifier architecture as a central building block is described in this paper. The complete system, which consists of a feature extraction module and the connectionist classifier module, has been designed and implemented in digital VLSl; system architectural aspects as well as the procedure of adaptation of the system to different segmentation problems are discussed. The performance of the segmentation system on real world problems is demonstrated using scenes from industrial inspection, texture recognition, and combustion chamber research tasks.
TL;DR: A system for segmentation and grouping, based on the pseudo-Wigner distribution a discrete approximation to the WD, is discussed, and experimental results are shown.
03 Jun 1991
03 Jun 1991
TL;DR: The Bayesian segmentation model developed is motivated by consideration of the information needed for higher-level visual processing, using the minimum description-length philosophy that the best segmentation allows the most efficient representation of visual data.
Abstract: The Bayesian segmentation model developed is motivated by consideration of the information needed for higher-level visual processing. A segmentation is regarded as a collection of parameters defining an image-valued stochastic process by separating topological (adjacency) and metric (shape) properties of the subdivision and intensity properties of each region. The prior selection is structured accordingly. The novel part of the representation, the subdivision topology, is assigned a prior by universal coding arguments, using the minimum description-length philosophy that the best segmentation allows the most efficient representation of visual data. >
TL;DR: Evaluation of the broad phonetic classification and segmentation of continuous speech on a multi-speaker database of continuously spoken number strings has demonstrated excellent segmentation and classification performances.
30 Aug 1991
TL;DR: The performance of the different moving object segmentation algorithms is evaluated, results are presented, and the image differenced algorithms presented are absolute differencing using an estimated reference, and temporal variance measurement.
Abstract: Moving object segmentation utilizes the temporal information inherent in a sequence of images to extract the moving objects from each image. Image registration, required in applications where the imaging sensor moves, quantifies the relative displacement between the images in the sequence. Three classes of image registration techniques are distinguished; correlation methods, Fourier methods, and feature point extraction. The moving object segmentation algorithms are divided into two categories. The image differencing algorithms presented are absolute differencing using an estimated reference, and temporal variance measurement. The selective segmentation algorithms presented are trajectory analysis, steadiness analysis, and shape analysis. The performance of the different moving object segmentation algorithms is evaluated and results are presented. >
Dissertation•
01 Mar 1991
TL;DR: It is suggested that within the context of the modern world, the use of icons and symbols to describe human interaction with the world around us is a natural extension of human contact with the Cosmos.
Abstract: ............................................... xii. PRINCIPAL ABBREVIATIONS AND SYMBOLS.................... xii
03 Jun 1991
TL;DR: A closed-loop image segmentation system that incorporates a genetic algorithm to adapt the segmentation process to changes in image characteristics caused by variable environmental conditions is presented.
Abstract: A closed-loop image segmentation system that incorporates a genetic algorithm to adapt the segmentation process to changes in image characteristics caused by variable environmental conditions is presented. The genetic algorithm efficiently searches the hyperspace of segmentation parameter combinations to determine the parameter set which maximizes the segmentation quality criteria. A summary of the experimental results that demonstrates the ability to perform adaptive image segmentation and to learn from experience using a collection of outdoor color imagery is given. >
14 Apr 1991
TL;DR: This method solves the problems of a segmentation-based image coding technique with constant segments by proposing a methodology for segmenting an image into texturally homogeneous regions with respect to the degree of roughness as perceived by the HVS.
Abstract: A new texture segmentation-based image coding technique which performs segmentation based on roughness of textural regions and properties of the human visual system (HVS) is presented. This method solves the problems of a segmentation-based image coding technique with constant segments by proposing a methodology for segmenting an image into texturally homogeneous regions with respect to the degree of roughness as perceived by the HVS. The segmentation is accomplished by thresholding the fractal dimension so that textural regions are classified into three texture classes: perceived constant intensity, smooth texture, and rough texture. An image coding system with high compression and good image quality is achieved by developing an efficient coding technique for each segment boundary and each texture class. Good quality reconstructed images are obtained with 0.08 to 0.3 b/p for three different types of imagery. >
TL;DR: This work has provided a quantitative measure for the partial evaluation of the segmentation which can be applied independent of attribute or combination of attributes.
Abstract: Radiological scans acquired using either the X-ray CT or the NMR imaging techniques' provide a wealth of information about tissue behaviour under that imaging modality and contrast agent. To reason about the image in an interpretation stage the scans have to be converted from a pixel by pixel representation to a symbolic form. The technique used by us to generate such a description is region-based segmentation. Each region refers to a pixel or group of pixels having a common attribute. This work has provided a quantitative measure for the partial evaluation of the segmentation which can be applied independent of attribute or combination of attributes. From our initial studies of the behaviour of CT scans a precept for segmentation was developed. The segmentation employs a one-to-one map as an adaptive mechanism. The segmentation criterion at each point in the image therefore depends on the value at the corresponding point in the map. Any process can be used to generate this map, and so easily utilizes new...
IBM1
TL;DR: A gray-level image segmentation method for use in segmentation-based image compression that consists of a variation of centroid-linkage region growing to perform the initial segmentation of the image, followed by nonlinear filtering to eliminate visually insignificant image segments.
Abstract: The authors describe a gray-level image segmentation method for use in segmentation-based image compression. The method consists of two steps: a variation of centroid-linkage region growing to perform the initial segmentation of the image, followed by nonlinear filtering to eliminate visually insignificant image segments. Both steps take advantage of human visual system properties to improve allocation of image segments. Subjective experiments have been conducted to determine the interactions and optimum balance between the steps. It is shown that the proposed two-step approach produces substantially better-quality segmented images than region growing used alone. >
01 Jul 1991
TL;DR: In this article, a graph theoretic approach to image segmentation is presented, and its application to tissue segmentation in MR images of the human brain is demonstrated, where an undirected adjacency graph G is used to represent the image with each vertex of G corresponding to a homogeneous component of the image.
Abstract: A novel graph theoretic approach to image segmentation is presented, and its application to tissue segmentation in MR images of the human brain is demonstrated. An undirected adjacency graph G is used to represent the image with each vertex of G corresponding to a homogeneous component of the image. Each component may be a single pixel or a connected region which, under a suitable criterion, is homogeneous. All pairs of nodes corresponding to spatially connected pixels or regions in the image are linked by arcs in G. A flow capacity, assigned to each arc, is chosen to reflect the probability that the pair of linked vertices belong to the same region or tissue type. The segmentation is achieved through clustering vertices in G by removing arcs of G to form mutually exclusive subgraphs. The subgraphs formed by the clustering algorithm are optimal in the sense that the largest inter-subgraph maximum flow is minimized. Each of the resulting subgraphs then represents a homogeneous region of the image. Using a suitable choice of the arc capacity function, this approach can be used to segment the image either by searching for statistically homogeneous regions (texture segmentation) or by searching for closed region boundaries (edge detection). A direct implementation of the new segmentation algorithm requires the construction of a flow equivalent spanning tree for G. As the size of the graph G increases, constructing an equivalent tree becomes very inefficient. In order to overcome this problem, an algorithm for hierarchically constructing and partitioning a partially equivalent tree of much reduced size has been developed. This hierarchical algorithm results in an optimal solution equivalent to that obtained by partitioning the complete equivalent tree of G.
03 Jun 1991
TL;DR: A non-supervised method of bayesian contextual radar image segmentation using the hierarchical image model (HIM) from Kelly and Derin in order to speed up the segmentation.
Abstract: We present a non-supervised method of bayesian contextual radar image segmentation. We adopt the hierarchical image model (HIM) from Kelly and Derin. In contrast to their global segmentation methcd a local method is used in order to speed up the segmentation. The algorithm for parameter estimation is SEM, a recent variation of EM. The algorithm obtained is tested on synthetic images and also applied to the segmentation of real SEAS AT-scenes.