Showing papers on "Orientation (computer vision) published in 1981"

A Computer Implementation of a Theory of Human Stereo Vision

Abstract: Recently, Marr & Poggio (1979) presented a theory of human stereo vision. An implementation of that theory is presented, and consists of five steps. (i) The left and right images are each filtered with masks of four sizes that increase with eccentricity; the shape of these masks is given by delta 2G, the Laplacian of a Gaussian function. (ii) Zero crossings in the filtered images are found along horizontal scan lines. (iii) For each mask size, matching takes place between zero crossings of the same sign and roughly the same orientation in the two images, for a range of disparities up to about the width of the mask's central region. Within this disparity range, it can be shown that false targets pose only a simple problem. (iv) The output of the wide masks can control vergence movements, thus causing small masks to come into correspondence. In this way, the matching process gradually moves from dealing with large disparities at a low resolution to dealing with small disparities at a high resolution. (v) When a correspondence is achieved, it is stored in a dynamic buffer, called the 2 1/2-dimensional sketch. To support the adequacy of the Marr-Poggio model of human stereo vision, the implementation was tested on a wide range of stereograms from the human stereopsis literature. The performance of the implementation is illustrated and compared with human perception. Also statistical assumptions made by Marr & Poggio are supported by comparison with statistics found in practice. Finally, the process of implementing the theory has led to the clarification and refinement of a number of details within the theory; these are discussed in detail.

A Representation for Visual Information.

Abstract: : This dissertation presents a new technique for representing digital pictures. The principal benefit of this representation is that it greatly simplifies the problem of finding the correspondence between components in the description of two pictures. This representation technique is based on a new class of reversible transforms (the Difference of Low Pass or DOLP transform). A fast algorithm for computing the DOLP transform is then presented. This algorithm, called cascade convolution with expansion is based on the auto-convolution scaling property of Gaussian functions. Techniques are then described for constructing a structural description of an image from its Sampled DOLP transform. The symbols in this description are detected by detecting local peaks and ridges in each band-pass image, and among all of the band-pass image. This description has the form of a tree of peaks, with the peaks interconnected by chains of symbols from the ridges. The tree of peaks has a structure which can be matched despite changes in size, orientation, or position of the gray scale shape that is described.

Parameter networks: towards a theory of low-level vision

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.

Parameter networks: Towards a theory of low level vision

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.

Defect analysis system

Abstract: Disclosed is an apparatus, and a related method, for the automatic detection of defects in manufactured parts, such as turbine blades and the like. The apparatus includes a video camera, a thresholding module to provide a binary image of a frame of video data, a region growing module to define anomalous regions by associating contiguous picture elements indicative of surface anomalies on the part, and defect analysis modules for determining if a rejectable defect exists, based on a comparison between statistics with respect to the anomalous regions and predetermined rejection criteria. Rejection may be based on region size, region shape, or region position and orientation with respect to neighboring regions. Thus, a region may be rejected if it is nearly collinear with another region, even though neither region may itself be characterized as a rejectable region.

The Image Irradiance Equation: Its Solution and Application

Abstract: : How much information about the shape of an object can be inferred from its image? In particular, can the shape of an object be reconstructed by measuring the light it reflects from points on its surface? These questions were raised by Horn (H070) who formulated a set of conditions such that the image formation can be described in terms of a first order partial differential equation, the image irradiance equation. In general, an image irradiance equation has infinitely many solutions. Thus constraints necessary to find a unique solution need to be identified. First we study the continuous image irradiance equation. It is demonstrated when and how the knowledge of the position of edges on a surface can be used to reconstruct the surface. Furthermore we show how much about the shape of a surface can be deduced from so called singular points. At these points the surface orientation is uniquely determined by the measured brightness. Then we investigate images in which certain types of silhouettes, which we call b-silhouettes, can be detected. In particular we answer the following question in the affirmative: Is there a set of constraints which assure that if an image irradiance equation has a solution, it is unique? To this end we postulate three constraints upon the image irradiance equation and prove that they are sufficient to uniquely reconstruct the surface from its image.

Acquisition of locational information about reference points during blindfolded and sighted locomotion: effects of a concurrent task and locomotion paths.

Abstract: Acquisition of cognitive maps (locational information about reference points) was investigated by requiring subjects to walk repeatedly according to paths consisting of 2, 4 and 6 linear segments. The starting point of each path was designated as a reference point, to which direction and distance were estimated by the subjects at the end of the path. Acquisition was indicated by the finding that variable errors and latencies decreased across blocks of trials. Interfering with central processing of locational information during locomotion by means of a concurrent task, while impairing orientation performance, did not seem to prevent acquisition. As suggested by an increase in latencies with the number of path segments, the subjects performing the concurrent task may have been able to infer and store the location of the starting point after walking. Finally, orientation performance did not improve for blindfolded subjects, probably because they were unable to use stored information about locations.

Measuring of feature for photo interpretation

Abstract: A method and apparatus is disclosed for determining the height, width, leh and orientation of an object from the feature image of the object found on oblique panoramic aerial photography. A magnifying eyepiece is provided with a calibrated reticle for measuring the dimensions of the feature image. The eyepiece also has a reference mark and indicia about the eyepiece for measuring the orientation from true north of the longitudinal axis of the feature image. The actual measurements of the object are quickly determined from the measurements of the feature image using a suitably programmed calculator supplied with the photograph and camera parameters.

Recovering intrinsic scene characteristics from images

Abstract: : The central theme of our research is the recovery of information about the three-dimensional structure and physical characteristics of surfaces depicted in an image -- their shapes, locations, and photometric properties. The main obstacle to surface recovery is the confounding of the desired properties in the sensory data: images are inherentlly ambiguous. Our approach to resolving this ambiguity rests on the application of generic, low-level knowledge (e.g., such basic assumptions as surface continuity and general position) to constrain the interpretation. The problem may be viewed as that of decomposing the image into its physically meaningful constituents -- surface orientation, reflectance, illumination, and so on. The 'intrinsic image model' provides a conceptual and computational framework in which this view is made explicit. Surface perception plays a fundamental role in early visual processing, both in humans and in machines. Work on surface perception has focused on the discrimination of edge types (e.g., extremal boundary or cast shadow), on the three-dimensional interpretation of edges, and on surface reconstruction by interpolating from edges and using texture geometry.

Video image simulation apparatus

Abstract: A video display system employs a plurality of controlled cathode ray tubes producing individual images which are combined by dichroic prisms to form a single image which is transmitted by a fiber-optic cable to a set of optical elements which transmit the image to the viewer. The position and orientation of the viewing instrument are sensed to provide control signals to an image generator to provide an image appropriate to the orientation of the viewing instrument.

Photogrammetric device for aircraft and spacecraft for producing a digital terrain representation

Abstract: A sensor device of the type to be carried on aircraft or spacecraft which avel along an arbitrary flight path, operates to provide data corresponding to the course and orientation of the craft, and a digital display of the terrain over which the craft is traveling. The device includes at least first, second and third sensor lines comprising rows of photosensitive semiconductor elements, arranged transversely of the flight path in an image plane. Each of the sensor lines provides line images corresponding to a terrain image projected onto the sensor line. At least one of the sensor lines extends in the image plane to form angles relative to the other sensor lines, so that the terrain line images produced by the angled sensor line intersect the terrain line images produced by the other sensor lines. This arrangement results in establishing clearly defined connecting points of the images, and provides high accuracy in the determination of the orientation data and of the terrain.

The Representation of Image Texture

Abstract: : This thesis explores how to represent image texture in order to obtain information about the geometry and structure of surfaces, with particular emphasis on locating surface discontinuities. Theoretical and psychophysical results lead to the following conclusions for the representation of image texture: (1) A texture edge primitive is needed to identify texture change contours, which are formed by an abrupt change in the 2-D organization of similar items in an image. The texture edge can be used for locating discontinuities in surface structure and surface geometry and for establishing motion correspondence; (2) Abrupt changes in attributes that vary with changing surface geometry -- orientation, density, length, and width -- should be used to identify discontinuities in surface geometry and surface structure; (3) Texture tokens are needed to separate the effects of different physical processes operating on a surface. They represent the local structure of the image texture. Their spatial variation can be used in the detection of texture discontinuities and texture gradients, and their temporal variation may be used for establishing motion correspondence. What precisely constitutes the texture tokens is unknown; it appears, however, that the intensity changes alone will not suffice, but local groupings of them may; and (4) The above primitives need to be assigned rapidly over a large range in an image. (Author)

Feature Extraction and Selection for Pattern Recognition of Two-Dimensional Hydrometeor Images

Abstract: New feature extraction techniques are developed for two-dimensional binary images of ice particles and raindrops. These features are employed in the statistical classification of these patterns into one of seven basic hydrometeor shapes. These images have been recorded by an airborne two-dimensional probe in order to provide information leading to an understanding of important physical processes in clouds. Minimum average probability of classification error is employed as the performance criterion with informal minimization of computing time. A synthetic image set was generated to develop feature extraction techniques. Moment normalization and rotation normalization are employed to convert all images to a common size and orientation. Ten time domain features are explored including a new development (circular deficiency) and a new application (cross correlation). Three frequency domain features are investigated including a new Fourier descriptor (centroid distance). An original method of reducing ...

Method and apparatus for identification of objects

Abstract: Disclosed is a method and an apparatus for identifying objects, such as articles sold in a store, appearing in a random position and orientation and at random times on an image window On a surface facing the image window, each object has an identification in the form of a field which comprises on at least one data track contrasting indicia with at least one contrasting line pattern identifying the position and the orientation of the data track The track also includes a plurality of parallel lines with variable spacing and/or line widths The image window is scanned line-by-line to generate a binary video signal which corresponds to the scanned contrast sequence The length of the overlapping light and dark intervals of the video signal resulting from the scanning of the line pattern is measured and successively measured interval lengths are compared with each other and a comparison signal having a first amplitude is generated when the two interval lengths which are being compared have a predetermined ratio to each other which conforms to the spacing of corresponding pattern lines An identification signal is emitted when during each of a number of successive comparison steps, the number of which is determined by the contrasting line pattern, a reference signal having the first amplitude is generated

Three-Dimensional Viewing Device For Examining Internal Structure

Abstract: A new 3-D viewing system will be described which allows a physician to simultaneously examine multiple CT or B-scan ultrasound scans in their proper orientation in all three dimensions. Several persons can view the display at the same time and there is no need for special glasses. Test images and line drawings displayed on a prototype viewing device exhibit both stereoscopic and motion parallax depth cues characteristic of a real three dimensional object. Operating principles of the device will be outlined and a new 3-D figure/ground illusion will be described.

The Implicit Constraints of the Primal Sketch.

Abstract: : Computational theories of structure-from-motion and stereo vision only specify the computation of three-dimensional surface information at points in the image at which the irradiance changes. Yet, the visual perception is clearly of complete surfaces, and this perception is consistent for different observers. Since mathematically the class of surfaces which could pass through the known boundary points provided by the stereo system is infinite and contains widely varying surfaces, the visual system must incorporate some additional constraints besides the known points in order to compute the complete surface. Using the image irradiance equation, we derive the surface consistency constraint, informally referred to as no news is good news. The constraint implies that the surface must agree with the information from stereo or motion correspondence. and not vary radically between these points. An explicit form of this surface consistency constraint is derived, by relating the probability of a zero-crossing in a region of the image to the variation in the local surface orientation of the surface, provided that the surface albedo and the illumination are roughly constant.

Extraction of Geological Lineaments from LANDSAT Imagery by Using Local Variance and Gradient Trend

Abstract: A method for computer extraction of geological lineaments from LANDSAT images is presented. The variance thresholded map of each image is separated into eight gradient direction maps, where basic linear segments are determined each of which is based on the principal eigen vector of a cluster of pixels whose gradient direction is orthogonal to its orientation. A special linking algorithm operates on the basic linear segments to give a candidate lineament segment. The method has been experimented on LANDSAT images of a small region southwest to Pittsburgh. The preliminary result is encouraging and further work for improvement is suggested.

Real-Time Three-Dimensional Vision For Parts Acquisition

Abstract: The National Bureau of Standards is developing a vision system for use in an automated factory environment. The emphasis of the project is on the real-time acquisition of three-dimensional parts using visual feedback. The system employs multiple light sources in con-junction with object models to establish the position and orientation of an object in the camera's field of view. A flood flash enables shape information to be obtained from an image, while a plane of light can be used to find the three-dimensional positions of points on the object. Because there are only a small number of object types and the objects all have pre-defined nominal locations, a model can be used to predict how the scene should look from a given viewpoint using a particular light source. This prediction can be compared with the actual image, and the differences used to establish position information. Models are expected to be particularly useful in reducing the number of views of an object necessary to calculate its three-dimensional position.

Sampling The Radon Transformation In Conventional Computed Tomography

Abstract: The sampling geometry for conventional, multidetector, computed tomography is illustrated in terms of the Radon transformation for both rotate-rotate (3rd generation fan beam) and rotate-stationary (4th generation fan beam) scanners. By deriving an expression for the outline of the sampling region in the Radon transformation for each detector measurement it is demonstrated that the entire Radon transformation can be covered by non-overlapping sampling regions with the assumption of negligible detector dead space. An expression for angular aliasing is derived which demonstrates that object dependent artifacts can occur if the angular width of the sampling regions in conjunction with the angular sampling increment does not provide sufficient suppression of the high order angular harmonics in the representation of the scanned object. The number of views necessary to suppress angular aliasing, as well as the potential spatial resolution and general image quality are shown to be fundamentally related to the size, shape, and relative orientation of the Radon transformation sampling regions.© (1981) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

Three-Dimensional Display Of Computed Tomographic Volume Images To Visualize Internal Organs

Abstract: Volume images made up of "stacks" of parallel computed tomographic (CT) cross-sectional images are displayed in three dimensions utilizing the method of projection imaging. This technique involves the mathematical projection of the volume picture elements (voxels) of the 3-D image onto a plane to form a two-dimensional projection image which, for x-ray CT volume images, resemble conventional radiographs. Projection images formed at two angles of view, 2° to 8° apart, are utilized as stereo-pair projections to view the volume image in three dimensions. Before projection, selected regions of the volume image are partially dissolved or totally removed from the volume to enhance the visibility of remaining struc-tures. These processes, referred to as numerical tissue "dissolution" and "dissection", are utilized to overcome the undesired effects of superposition which occur as natural consequence of displaying a stack of cross sections as a volume image, i.e., deeper image regions are obscured by overlying structure. Examples are shown where overlying regions of the volume image have been "cut" from the volume to more clearly visualize deeper anatomy. Particular emphasis is given to the use of these methods in identifying two-and three-dimensional subregions of interest within the volume for further detailed view-ing and quantitative analysis. As an example, the use of the 3-D display of volume images to guide the process of identifying the optimal orientation of oblique section images through internal organs of the body is illustrated.© (1981) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

Quantitative methods applied to stereo imaging. Theory.

Abstract: Publisher Summary This chapter describes the theory of coordinate systems and transformation to stereo imaging. The term quantitative refers to precise dimensional measurements made to obtain direct information on the object being measured or derived information. The process of measuring two-dimensional photo-coordinates requires a reference system established with a set of marks that can be observed on each photograph and that would maintain a fixed relationship to the imaging system. The model-coordinate system refers to the stereoscopic model, established analogically at a stereo instrument, or mathematically by means of a computer or calculator. This coordinate system is three-dimensional, orthogonal, and right-handed. The location of the origin and the orientation of the system may be arbitrary, depending on the convenience of the use of the EM data. In practice, the X′ direction coincides with the base line or the direction of tilt used in creating the stereo model. The object-coordinate system is the final three-dimensional system of reference to the object. It is also right-handed and orthogonal. The principal features of the perspective projection are a perspective center and finite projection distance.

Automatical Contour Production Using Image Correlation of Aerial Stereo Photographs—Processing and Automatic Orientation

Abstract: 走査型微少濃度計と汎用電子計算機を使用して, 実体空中写真から数値地形モデルおよび等高線を自動的に製作する1つの方法について一連の実験を行った。この方法は基本的にはKreilingの研究とKonecnyの提案に基づいているが実験によっていくつかの新しい成果を得た。この論文は空中写真の画像分解から自動相互標定までを実験に則して説明した。まず空中写真をドラム走査型微少濃度計で画像分解し, 画像の行, 列を軸とする座標系を4隅の指標を使って較正する。つぎに左画像に与えた標定点を, 画像の相関を使って右画像の中で探索し, 解析的に標定要素を求めた。これに続く処理は左右画像を偏位修正してファイル処理を効率的に行えるようにした後, 地上座標系の格子点標高を順次求めていく手続きをとっているがこれについては別に報告する。前半の実験から得た注目すべき結果としては,1) 現在のドラム走査型微少濃度計の機構ではフィルムの変形による画像の不規則歪みがかなり大きく, 写真測量の観点からは許容度を越える。2) 標定の自動化については, コントラストの高い点を標定点として左画像に与えればその対応点を機械的に捜し出すことができ, 相互標定の結果も目測でグレイマップを使って行った位置同定より高い精度が得られた。3) しかし任意に標定点を与えた場合には, 山地にある対応点を捜し出すことはできなかった。

Visual tracking in stereo

Abstract: A method is described for visual object tracking by a computer vision system using TV cameras and special low-level image processing hardware. The tracker maintains an internal model of the location, orientation, and velocity of the object in three-dimensional space. This model is used to predict where features of the object will lie on the two-dimensional images produced by stereo TV cameras. The differences in the locations of features in the two-dimensional images as predicted by the internal model and as actually seen create an error signal in the two-dimensional representation. This is multiplied by a generalized inverse Jacobian matrix to deduce the error in the internal model. The procedure repeats to update the internal model of the object's location, orientation and velocity continuously.

Modeling 3-D Structure

Abstract: Recognition of 3-D objects and the determination of their orientation in space are two major problems of robot vision systems. Moreover, in an industrial environment, these tasks should be performed quickly and accurately. A simple representation of 3-D objects is given which makes possible a technique for recognition and orientation determination of 3-D objects in laser range images. This technique is an extension of the 2-D Hough shape transform to handle 3-D surfaces; the technique is applied directly to a set of 3-D points extracted from a range image.

Applying photogrammetry to real time collection of digital image data

Abstract: : The practicality of converting digital and stereo image data into terrain elevation data in real time is investigated. Results indicated that the real time collection of elevation data using epipolar correlation methods is not feasible. It is shown that real time mensuration of 7-parallax can be used as a means to help maintain a desired relative orientation in a variety of digital collection system. (Author)