Abstract: One of the most fundamental problems in vision is segmentation; the way in which parts of an image are perceived as a meaningful whole. Recent work has shown how to calculate images of physical parameters from raw intensity data. Such images are known as intrinsic images, and examples are images of velocity (optical flow), surface orientation, occluding contour, and disparity. While intrinsic images are not segmented, they are distinctly easier to segment than the original intensity image. Segments can be detected by a general Hough transform technique. Networks of feature parameters are appended to the intrinsic image organization. Then the intrinsic image points are mapped into these networks. This mapping will be many-to-one onto parameter values that represent segments. This basic method is extended into a general representation and control technique with the addition of three main ideas: abstraction levels; sequential search; and tight counting These ideas are a nucleus of a connectionist theory of low 'eve and m'ermediate-level vision. This theory explains segmentation in terms of massively parallel cooperative computation among intrinsic images and a set of parameter spaces at different levels of abstraction.

Abstract: One of the most fundamental problems in vision is segmentation; the way in which parts of an image are perceived as a meaningful whole. Recent work has shown how to calculate images of physical parameters from raw intensity data. Such images are known as intrinsic images, and examples are images of velocity (optical flow), surface orientation, occluding contour, and disparity. While intrinsic images are not segmented, they are distinctly easier to segment than the original intensity image. Segments can be detected by a general Hough transform technique. Networks of feature parameters are appended to the intrinsic image organization. Then the intrinsic image points are mapped into these networks. This mapping will be many-to-one onto parameter values that represent segments. This basic method is extended into a general representation and control technique with the addition of three main ideas: abstraction levels; sequential search; and tight counting These ideas are a nucleus of a connectionist theory of low 'eve and m'ermediate-level vision. This theory explains segmentation in terms of massively parallel cooperative computation among intrinsic images and a set of parameter spaces at different levels of abstraction.

Abstract: A fast nonlinear time alignment method is presented, which is based on a preprocessing of the normalized speech spectrogram by means of a segmentation of the trace in the spectral feature space. After such trace segmentation the patterns have a fixed format and allow for a subsequent classification with a distance measure which is obtained from conventional dynamic programming with extreme constraints. Since, due to the trace segmentation preprocessing, these extreme constraints can be applied without performance degradation, the described method offers savings in computing time by a factor of 10 or more as compared to conventional dynamic programming. As a side benefit, reference pattern memory savings by a factor of 3 or more are obtained.

Abstract: Speech is modeled as a Markov chain. Scoring is developed to convert observations of the speech signal into estimated probabilities of the locations of segment boundaries. Dynamic programming is then used to compute a most‐probable segmentation for the speech. The process automatically adjusts to speakers and incorporates a priori information in a probabilistic and systematic fashion. The performance of the algorithm appears to be state‐of‐the‐art, independent of speaker.

Abstract: : New results are presented in the areas of shape matching of nonoccluded and occluded objects in two dimensions, surface approximation by polygons, shape matching of objects in three dimensions, and segmentation of images having unimodal distributions. The same stochastic labeling technique is used in both shape matching and segmentation with various extensions. Shape matching is viewed as a segment matching problem. Unlike the previous work in shape matching of 2-D objects, the technique is based on a stochastic labeling procedure which explicitly maximizes a criterion function based on the ambiguity and inconsistency of classification. To reduce the computation time, the technique is hierarchical and uses results obtained at low levels to speed up and improve the accuracy of results at higher levels. This basic technique has been extended to the situation where various objects partially occlude each other to form an apparent object and our interest is to find all the objects participating in the occlusion. In such a case several hierarchical processes are executed in parallel for every participating object in the occlusion and are coordinated in such a way that the same segment of the apparent object is not matched to the segments of different actual objects. These techniques have been applied to two-dimensional shapes represented by polygons and the power of the techniques is demonstrated by the examples taken from synthetic, aerial, industrial and microscope images, where the matching is done after using the actual segmentation methods.

Abstract: The efficiency of syllabic segmentation and recognition is demonstrated in an experiment using three different word recognition systems and a vocabulary of 1000 words. In each system the preprocessing is carried out by a special loudness analyzer which yields 22 specific loudness functions. The first system avoids any segmentation and the total word pattern is time normalized to a constant length. In the second system syllable nuclei are detected and used as segment boundaries; the segments are time normalized and the resulting word pattern classified. The third system classifies each demisyllable using vowels and consonant clusters as decision units. For small vocabularies the first system gives the best performance. For more than 500 words the performance of the second system with syllabic segmentation surpasses that of the first system. For this vocabulary size however, the demisyllable recognition performs best and is significantly advantageous for a vocabulary consisting of 1000 words.

Abstract: A remote sensing user does not photointerprete image pixels, but entities. Therefore, there is a segmentation processing, previous to the recognition itself. What we propose in this paper, is to automate the segmentation by using of monospectral, multispectral and multitemporal properties, measured by several criteria. The combination of these criteria is performed by means of tools of the fuzzy sets theory. A designated entity is automatically segmented by combining a sequence of criteria in order to converge towards the final decision without any thresholding, weighing, ..• The ready access to the multi temporal data belonging to a same designated entity, is obtained by comparing the segmentation results at different dates, through geometric deformation models. Finally the radiometries, extracted entity/ entity, by using this segmentation method, feed the diachronic analysis in the context of the Lauragais experiment.

Abstract: This paper presents results of experimental studies in image understanding. Two experiments are discussed, one on image correlation and another on target boundary estimation. The experiments are demonstrative of polar exponential grid (PEG) representation, an approach to sensory data coding which the authors believe will facilitate problems in 3-D machine perception. Our discussion of the image correlation experiment is largely an exposition of the PEG-representation concept and approaches to its computer implementation. Our presentation of the boundary finding experiment introduces a new robust stochastic, parallel computation segmentation algorithm, the PEG-Parallel Hierarchical Ripple Filter (PEG-PHRF).© (1981) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

Abstract: Recognition of connected word strings can be performed by segmenting the word string automatically into single-word components which are then classified by a single-word recognition system. We propose and investigate a segmentation procedure which is based completely on statistical principles. An estimation algorithm, adapted to the statistical data of the signal parameters, determines the word boundaries. This procedure, which offers several advantages over other methods, has been tested with connected digits. The results show that an estimation algorithm based on quadratic polynomials yields sufficiently accurate segmentation. Recognition results for 2-to 4-digit strings are presented in this paper.

Abstract: In visualizing objects in 3-D digital images based on shaded-surface displays, segmen-tation of the 3-D regions corresponding to the objects to be visualized is a basic processing operation. When object and non-object regions touch or overlap and have similar features, automatic segmentation methods fail. To overcome this problem, a method, based on creating an arbitrarily-shaped three-dimensional window that encloses only the object regions, is proposed in this paper. The window is created using the contours traced in a series of 'slices' comprising the 3-D image. The capability to create arbitrarily-shaped 3-D windows also permits 3-D visualization corresponding to arbitrary intersections of the object. Having identified the object regions in the 3-D image, the next important processing operation before the 3-D display of an object can be achieved is the formation of its boundary surfaces. A simple and fast algorithm to form boundary surfaces from a given set of contours is presented in this paper which makes use of the 'cuberille model' for the representation of discrete surfaces. The performance of the algorithms is illustrated using computerized tomography data.

Abstract: In order to speed up the segmentation of image sequences containing moving objects, an approach is proposed which employs knowledge from different sources to control the extraction of regions. Here, regions are connected image components which exhibit grayvalue characteristics determined by specific values of the parameters supplied to the extraction algorithm. These values may be inferred from outside knowledge and previous interpretation results, e.g. from older frames in a sequence. Various segmentations of the same image with parameters adapted to different expected object surface characteristics may be obtained. Thus, this approach is an attempt to overcome the sharp partition of segmentation and interpretation found in conventional scene analysis paradigms.

Abstract: Implementation of a three-dimensional video tracking algorithm using CCD's is presented. The algorithm does not require conventional pattern recognition or continuous segmentation updating. The only requirements, for the class of targets presented, are that the sampling lattice contain only points inside the target and that the background vary slowly. Several well-known segmentation techniques are available for initial segmentation and determination of target points. The use of CCD devices simplifies the architecture of the processor and permits real time processing at a high sampling rate. Some of the vector and matrix operations require either the use of continuously variable tapweight transversal filters or analog multipliers. The use of analog multipliers, which can be on-chip, reduces the number of components and offers higher accuracy than presently available variable tap weight transversal filters.

Abstract: Segmentation in the field of optical character recognition involves forming vertical histograms from video data representing a line of characters, for example counts of black picture elements at imaginary vertical lines along the line of characters. The vertical histograms are used in a variety of ways e.g. in segmenting characters after location of the baseline using horizontal histograms, in segmenting words in the line, and in determining underscore, such underscores being masked during subsequent segmentation and recognition of the underscored character.

Abstract: Picture segmentation is expressed as a sequence of decision problems within the framework of a split-and-merge algorithm. First regions of an arbitrary initial segmentation are tested for uniformity and if not uniform they are subdivided into smaller regions, or set aside if their size is below a given threshold. Next regions classified as uniform are subject to a cluster analysis to identify similar types which are merged. At this point there exist reliable estimates of the parameters of the random field of each type of region and they are used to classify some of the remaining small regions. Any regions remaining after this step are considered part of a boundary ambiguity zone. The location of the boundary is estimated then by interpolation between the existing uniform regions. Experimental results on artificial pictures are also included.

Abstract: In this paper phonemic segmentation methods of speech are studied in which the segmentation is performed by using the extrema of a time-varying signal. In most cases the number of extrema in the segmentation signal and thus also the number of characters in the phonemic transcription tend to be too large. A new method which utilizes the time behaviour of the segmentation signal and the phonemic labeling result is introduced to decrease the number of phonemic characters. In the practical experiments the performance of the proposed method is evaluated and a comparison of three segmentation methods is presented.

Abstract: Zadeh (4) introduced the notion of fuzzy set theory and Goetcherian (6) showed how fuzzy logic could extend a variety of binary image processes into a multi-valued (grey image) domain. This paper introduces the basics of fuzzy set theory and discusses the suitability of the physically parallel CLIP machine for implementing the multi-valued operators. A form of spatial filtering is explained and its application to shape discrimination and image segmentation is shown. Finally, an application to a real image processing problem is illustrated.

Abstract: An important and difficult problem in image processing is to segment an image into several different kinds of regions. Unless we isolate regions, we cannot perform further processing for pattern recognition, scene analysis, image interpretation, etc. Besides, if the segmentation phase can be done at high speed, this reduces the total processing time greatly.

Abstract: PURPOSE:To improve a segmentation processing efficiency, by executing the designating of a line pitch including a free pitch and the designation of a character pitch including a free pitch independently of each other for every form. CONSTITUTION:A control part 7 confirms a line projection register 3 to monitor presence or absence of a pattern in every time that subscanning advances for a constant length shorter than the minimum character height in a photoelectric conversion part 2. When line projection comes to the center of register 3, the control part 7 cuts out the first line from a buffer 5 in this position at, for example, the 1/2 inch per line. In parallel with this operation, the control part 7 monitors a column projection register 4 to cut out a character pattern from the buffer 5 at prescribed character pitch 1/4 inch in the position where column projection seemed to correspond to the first character comes to the center of the register 4. Hereafter, respective characters are cut out on a basis of the first character and are transferred to a recognizing part 6.

Abstract: PURPOSE:To recognize symbol patterns differing in size without thinning-out processing by extracting features of segmentation areas differing in size and by checking whether the amount of features varies with the size of the segmentation area CONSTITUTION:Video data of an input pattern is read out of plane memory 23 by control part 24 and when the area to be read reaches each segmentation areas of 44X44, 28X28 and 20X20 meshes by area setting parts 28, 29 and 30, a signal of logic ''1'' is generated and written in corresponding buffer memories 25, 26 and 27 Then, the contents of memory 25 are read successively and led to memories 41-44 corresponding to four sub areas of memories 45 and 46 and 44X44 segmentation area and the contents of memories 26 and 27 are supplied to extraction parts 45 and 46 in time-division mode Consequently, closed loops differing in size and symbol patterns of different internal structure are discriminately recognized by checking whether the amount of features varies with the size of the segmentation area