Showing papers on "3D reconstruction published in 1995"

Object-centered surface reconstruction: combining multi-image stereo and shading

Abstract: Our goal is to reconstruct both the shape and reflectance properties of surfaces from multiple images. We argue that an object-centered representation is most appropriate for this purpose because it naturally accommodates multiple sources of data, multiple images (including motion sequences of a rigid object), and self-occlusions. We then present a specific object-centered reconstruction method and its implementation. The method begins with an initial estimate of surface shape provided, for example, by triangulating the result of conventional stereo. The surface shape and reflectance properties are then iteratively adjusted to minimize an objective function that combines information from multiple input images. The objective function is a weighted sum of stereo, shading, and smoothness components, where the weight varies over the surface. For example, the stereo component is weighted more strongly where the surface projects onto highly textured areas in the images, and less strongly otherwise. Thus, each component has its greatest influence where its accuracy is likely to be greatest. Experimental results on both synthetic and real images are presented.

Relative 3D reconstruction using multiple uncalibrated images

Abstract: In this article we show how relative 3D reconstruction from point correspondences of multiple uncalibrated images can be achieved through reference points. The original contributions with respect to related works in the field are mainly a direct global method for relative 3D reconstruction and a geometric method to select a correct set of reference points among all im age points. Experimental results from both simulated and real image sequences are presented, and robustness of the method and reconstruction precision of the results are discussed.

Scaled Euclidean 3D reconstruction based on externally uncalibrated cameras

Abstract: Previous work shows that based on five non-coplanar correspondences of two uncalibrated cameras, 3D reconstruction can be achieved under projectile models, or based on four non-coplanar correspondences of two uncalibrated cameras, 3D reconstruction can be achieved under affine models, with three unknown parameters. In this paper, we show that based on four coplanar correspondences of two externally uncalibrated cameras, 3D reconstruction can be achieved in Euclidean space with only one unknown scaling parameter. Moreover, the unknown scale factor is the physical distance from the camera center to the plane formed by the four points in 3D space. If this distance is known a priori, then the 3D structure can be completely recovered. Both simulated and real data experimental results show that our reconstruction algorithm works reasonably robustly.

A system for the 3D reconstruction of retracted-septa PET data using the EM algorithm

Abstract: We have implemented the EM reconstruction algorithm for volume acquisition from current generation retracted-septa PET scanners. Although the software was designed for a GE Advance scanner, it is easily adaptable to other 3D scanners. The reconstruction software was written for an Intel iPSC/860 parallel computer with 128 compute nodes. Running on 32 processors, the algorithm requires approximately 55 minutes per iteration to reconstruct a 128/spl times/128/spl times/35 image. No projection data compression schemes or other approximations were used in the implementation. Extensive use of EM system matrix (C/sub ij/) symmetries (including the 8-fold in-plane symmetries, 2-fold axial symmetries, and axial parallel line redundancies) reduces the storage cost by a factor of 188. The parallel algorithm operates on distributed projection data which are decomposed by base-symmetry angles. Symmetry operators copy and index the C/sub ij/ chord to the form required for the particular symmetry. The use of asynchronous reads, lookup tables, and optimized image indexing improves computational performance. >

Ultrasonic three-dimensional reconstruction of the heart.

Abstract: The recent advances in ultrasound equipment, digital image acquisition, and display techniques made three-dimensional (3D) echocardiography a clinically feasible and exciting technique which allows objective analysis of structure and pathological conditions of complex geometry. In this report, different image acquisition techniques are described and compared. In our experience, with rotational scanning the acquisition of cross-sections for 3D reconstruction becomes an integral part of a routine diagnostic study, both with a multiplane transesophageal imaging transducer, and in precordial echocardiography. After digital reformatting and image processing, a volumetric data set is obtained, which allows the display of synthetic cross-sections in various orientations independent from the point of origin of the sector scan [anyplane two-dimensional (2D) imaging]. This also offers the possibility of volume quantification, without the assumption of theoretical geometrical model of the cavity. Finally, dynamic volume rendered display can be applied for the objective display of the anatomy and the complex relationship among the different structures.

Design and implementation of a database and program for 3D reconstruction from serial sections: a data-driven approach.

Abstract: This paper describes a structural approach for a standard setup of a computer program for 3D reconstruction from serial sections. Three-dimensional reconstruction as a technique increases in importance as, along with modern immunohistochemical techniques, it is a tool in the understanding of three-dimensional development patterns. In order to apply 3D reconstruction technique in a standard laboratory setup, an attempt was made to streamline the input and the manipulation of the data such that results are obtained easily. One will find a combination of two approaches in this paper: the first is a strict ordering of the complex data, and the second is an ordering of the processes that one wishes to apply on the data (together, these two approaches constitute an information analysis); because it was observed that developmental biologists tend to work from simple lines to describe their observations, the contour model was chosen as the vehicle to build a reconstruction model from. Consequently, the data is ordered in a database that has to be manipulated to get the data out in the desired format. The most important output format is a display of the reconstructed contour stack on a graphical computer screen. Together with the other data manipulation processes, such as the input, the inspection, the revision (correction), and the reconstruction, all processes are described using the reconstruction of an 11 embryonic days (ED) rat embryo as an example. Finally, the merits of the program are illustrated with an example from the development of the human embryonic heart.

Weakly-calibrated stereo perception for rover navigation

Abstract: Presents a vision system for autonomous navigation based on stereo perception without 3D reconstruction. This approach uses weakly calibrated stereo images, i.e. images for which only the epipolar geometry is known. The vision system first rectifies the images, matches selected points between the two images, and then computes the relative elevation of the points relative to a reference plane as well as the images of their projections on this plane. We have integrated this vision module into a complete navigation system. In this system, the relative elevation is used as a shape indicator in order to compute appropriate steering directions everytime a new stereo pair is processed. We have conducted initial experiments in unstructured, outdoor environments with an wheeled rover. >

Epipolar reconstruction of 3D structures

Abstract: A method of reconstructing selected features of a part manufactured according to a CAD model involves acquiring a set of linear push broom (LPB) projection images of the part acquired at different angles about an axis of rotation passing through the part. Acquiring a set of matrices M j j=1 . . . N, each of which maps 3D coordinates of the part to screen coordinates of one of the projection images. Reconstruction of 3D structures from the projection images requires identification of screen coordinates of each image which correspond to a point of the structure of the part to be reconstructed. Back projecting these screen coordinates modified by the distortion inherent in the LPB imaging device. This is accomplished by selecting a screen coordinate on a feature desired to be reconstructed. Computing a ray passing through the selected screen coordinate through an imaging center. Using each M matrix to map this ray to a hyperbola on the other images. The hyperbola is then scanned until the desired feature is encountered. This second screen location of this projection image corresponds to the first selected screen location. A second ray through the second screen coordinate is constructed. Where it intersects the first ray identifies the 3D reconstruction point. This is repeated for a number of points to create a reconstruction of the selected feature.

A Virtual Reality Medical Training System

Abstract: This paper presents the result of the interdisciplinary cooperation of traumatologists of the Berufsgenossenschaftliche Unfallklinik (BGU) in Frankfurt am Main and a team of computer graphics scientists of the Fraunhofer Institute for Computer Graphics in Darmstadt. We have developed a highly interactive medical training simulator system for arthroscopy by means of computer graphics and Virtual Reality techniques. This system offers an effective alternative to conventional training systems for training and establishing of arthroscopie techniques. The two main development tasks will be discussed: the 3D reconstruction process and the 3D interaction.

3D reconstruction in electrical impedance imaging using a direct sensitivity matrix approach

Abstract: This paper presents a reconstruction algorithm using a direct sensitivity matrix (DSM) approach for fast 3D image reconstruction in electrical impedance imaging. The boundary element method (BEM) is used in the construction of this matrix. The first images of a conductivity perturbation inside a sphere are reconstructed, using theoretical data.

A Generic Approach for Planar Patches Stereo Reconstruction

Abstract: Whereas edge-based stereo has traditionally received a lot of attention, relatively few explicit region-based 3D reconstruction techniques have been developed, despite the significant advantages such global features present for representing indoor or urban environments. We propose a general framework for 3D reconstruction of planar patches from stereoscopic pairs of images, and explicit a generic planar equation recovery scheme, which can then be applied to several types of stereosystem geometry. Another advantage of the formulation is its possible use on various cues: we propose three different applications, using moments of inertia, parametric and non-parametric photometric analysis. In each case, the robustness of the results is assessed, both with analytic data consistency check and tests performed on synthetic stereograms. Experiments on real data are also presented.

Robust computational vision

Abstract: This paper presents a paradigm for formulating reliable machine vision algorithms using methods from robust statistics. Machine vision is the process of estimating features from images by fitting a model to visual data. Computer graphics programs can produce realistic renderings of artificial scenes, so our understanding of image formation must be quite good. We have good models for visual phenomena, but can do a better job of applying the models to images. Vision computations must be robust to the kinds of errors that occur in visual signals. This paper argues that vision algorithms should be formulated using robust regression methods. The nature of errors in visual signals will be discussed, and a prescription for formulating robust algorithms will be described. To illustrate the concepts, robust methods have been applied to three problems: surface reconstruction, image flow estimation, and dynamic stereo.

Imaging diffusive media using time-independent and time-harmonic sources: dependence of image quality on imaging algorithms, target volume, weight matrix, and view angles

Abstract: We present results examining the dependence of image quality for imaging in dense scattering media as influenced by the choice of parameters pertaining to the physical measurement and factors influencing the efficiency of the computation. The former includes the density of the weight matrix as affected by the target volume, view angle, and source condition. The latter includes the density of the weight matrix and type of algorithm used. These were examined by solving a one-step linear perturbation equation derived from the transport equation using three different algorithms: POCS, CGD, and SART algorithms with contraints. THe above were explored by evaluating four different 3D cylindrical phantom media: a homogeneous medium, an media containing a single black rod on the axis, a single black rod parallel to the axis, and thirteen black rods arrayed in the shape of an 'X'. Solutions to the forward problem were computed using Monte Carlo methods for an impulse source, from which was calculated time- independent and time harmonic detector responses. The influence of target volume on image quality and computational efficiency was studied by computing solution to three types of reconstructions: 1) 3D reconstruction, which considered each voxel individually, 2) 2D reconstruction, which assumed that symmetry along the cylinder axis was know a proiri, 3) 2D limited reconstruction, which assumed that only those voxels in the plane of the detectors contribute information to the detecot readings. The effect of view angle was explored by comparing computed images obtained from a single source, whose position was varied, as well as for the type of tomographic measurement scheme used (i.e., radial scan versus transaxial scan). The former condition was also examined for the dependence of the above on choice of source condition [ i.e., cw (2D reconstructions) versus time-harmonic (2D limited reconstructions) source]. The efficiency of the computational effort was explored, principally, by conducting a weight matrix 'threshold titration' study. This involved computing the ratio of each matrix element to the maximum element of its row and setting this to zero if the ratio was less than a preselected threshold. Results obtained showed that all three types of reconstructions provided good image quality. The 3D reconstruction outperformed the other two reconstructions. The time required for 2D and 2D limited reconstruction is much less (< 10%) than that for the 3D reconstruction. The 'threshold titration' study shows that artifacts were present when the threshold was 5% or higher, and no significant differences of image quality were observed when the thresholds were less tha 1%, in which case 38% (21,849 of 57,600) of the total weight elements were set to zero. Restricting the view angle produced degradation in image quality, but, in all cases, clearly recognizable images were obtained.

Image intensifier-based volume tomographic angiography imaging system: system evaluation

Abstract: An image intensifier-based rotational volume tomographic angiography imaging system has been constructed. The system consists of an x-ray tube and an image intensifier that are separately mounted on a gantry. This system uses an image intensifier coupled to a charge coupled device (CCD) camera as a two-dimensional detector so that a set of two-dimensional projections can be acquired for a direct three-dimensional reconstruction (3D). A computer simulation has been performed to quantify the reconstruction error due to using 'a single circle cone beam geometry.' The results from the computer simulation indicate that the reconstruction error is relatively small compared to object contrast encountered in intra- arterial angiography. In this study, a scatter and veiling glare correction technique is reported. This technique put a beam stop array behind a phantom and in front of the image intensifier to estimate the scatter and veiling glare distribution. Then the primary beam distribution was estimated by subtracting the scatter and veiling glare distribution from the total detected radiation distribution. After obtaining the primary projections, a direct 3D reconstruction was obtained. The experimental results indicate that image contrast can be substantially improved by applying our scatter and veiling glare correction technique.

Surface Reconstruction Based on Visual Information

Abstract: Computer Vision Group, FR 3-11 Computer Science Department, Berlin Technical University, D-10587 Berlin, Germany http://www.cs.tu-berlin.de/~cvworld {klette,koschan,karsten,vr}@cs.tu-berlin.de Abstract The computation of surface data based on visual information is an important sub-component in the computer-graphical surface reconstruction of solids and in the control of 3-D environments. Different methodologies can be used for that as, e.g., static stereo, shape from motion, shape from shading, photometric stereo, or structured lightening. There exist different basic approaches in literature often based on simplifying assumptions. However, it is well known that such assumptions may not be true if surface reconstruction is applied under practical circumstances. In this paper, several problems are mentioned which are related to practical applications of surface reconstruction approaches following the methodologies of static stereo, shape from motion, and photometric stereo. We present specific solutions to cope with these problems, or the solution state what was reachable in our work. Some problems are ill-posed and limitations of approaches have to be accepted. As a second contribution of this paper, we discuss the evaluation problem of surface reconstruction algorithms. It is important to answer such questions as 1) For what kind of surfaces and 3-D objects an algorithm behaves either well or bad? 2) How accurate are the reconstruction results of an algorithm under specified circumstances? What measure can be used to evaluate reconstruction accuracy? 3) How to compare reconstruction results following different methodologies? 4) What algorithm can be suggested for a specific application project? and so on. So far we present some proposals and first quantitative or qualitative results for answering such questions. In our opinion a methodology for evaluating surface reconstruction algorithms is still at its beginning. However a critical evaluation of potential methods in project applications is helpful in selecting the appropriate algorithm.

Sub-pixel reconstruction of a variable albedo Lambertian surface

Abstract: Using a probabilistic interpretation of an n dimensional extension of Papoulis's Generalized Sampling Theorem, an iterative algorithm has been devised for 3D reconstruction of a Lambertian surface at subpixel accuracy. The problem has been formulated as an optimization one in a Bayesian framework. The latter allows for introducing a priori information on the solution, using Markov Random Fields (MRF). The estimated 3D features of the surface are the albedo and the height which are obtained simultaneously using a set of low resolution images. keywords: 3D Super resolution, Generalized Sampling Expansion, Low level image processing, Markov Random Fields (MRF).

Estimation of search-space in 3D coronary artery reconstruction using angiographic biplane images

Abstract: 3D reconstruction of heart arteries is of great interest, not only for diagnostic purposes, but also at a therapeutic level. 3D reconstruction, based on two radiographic images, is a multiple solution problem. The correct solution can only be obtained with a priori knowledge. Up to now, the number of reconstruction solutions, which depends on the image set, has not been studied. Almost all the reconstruction methodologies proposed up to now, try to avoid searching the total solution space. The present research try to determine the number of solutions (size of the solutions space), based on the generation of all the possible reconstruction solutions. This methodology implies the reconstruction of all the possible 3D segments and a selection of subsets that represents possible solutions. A first implementation validated using real data is discussed.

3D Reconstruction of Blood Vessels by Multi-Modality Data Fusion Using Fuzzy and Markovian Modelling

Abstract: We propose an original approach for 3D reconstruction of blood vessels based on fusion of digital angiography and intravascular echography data, without any geometrical a priori vessel model. We aim at better understanding and interpreting vessel morphology and atheromateous vascular lesions. A Markovian and a fuzzy approach are developed, which take into account all information about the problem, including its imprecisions. They both show good and similar results.

Stereographic viewing of 3D ultrasound images: a novelty or a tool?

Abstract: Three dimensional (3D) reconstruction of multiple two dimensional images, spatially and time registered, is increasingly being employed. These can be visualized on computer screens in full stereographic perspective using software and stereographic glasses now available. The authors have applied stereographic viewing to over 40 reconstructions of outlines of the endo and epicardiums of ultrasound imaged hearts. They found that standard viewing often mislead them into believing that misregistration existed. However, when viewed in stereographic mode, they found that these suspected lines did intersect, but at other places in the 3D reconstruction. The stereographic method has proven very useful in this and in distinguishing the separability of various 3D structural features in grid, surface, and voxel echocardiogram reconstructions.

Euclidean 3D Reconstruction from Stereo Sequences with Variable Focal Lengths

Abstract: A stereo rig can be calibrated using a calibration grid, but recent work demonstrated the possibility of auto-calibration. There remain two important limitations, however. First, the focal lengths of the cameras should remain fixed, thereby excluding zooming or focusing. Second, the stereo rig must not purely translate, which however is the most natural type of motion. This also implies that these methods collapse when the motion comes close to being a translation.

Brain surface scanning using structured light

Abstract: A significant problem in 3D reconstruction of biological tissue from histological material is alignment of the individual sections. We are developing a method to determine the surface of the tissue prior to cryosectioning and then utilize that information to guide registration. Toward that end, we have developed a structured light techniuqe for imaging frozen rat brains. The imaging approach relies on a novel coding scheme for the projected light which is based on 2D perfect submaps. Perhaps submaps are r by v c-ary arrays in which every n by m c-ary submatrix is unique. This coding scheme offers two major advantages over previous structured light patterns critical in the present application. It permits rapid image capture and, because each subwindow is unique, is robust in the presence of partial occlusion. To examine the accuracy of this technique, we compare the points mapped using it to the surface produced by block-face imaging. In the later approach, the tissue block is imaged prior to collecting each of the tissue sections. Since the block can be accurately repositioned after each cutting stroke, reconstruction of the surface from the block-face images is straightforward.

Measuring the shape of time-varying objects

Abstract: This thesis describes the first implementation of a colour encoded structured light (CESL) range-finder capable of measuring the shape of time-varying or moving surfaces. The system is shown to have mean square accuracies of better than 0.5mm when measuring the shape of the human mouth during continuous speech sampled at 50Hz, and better than 0.2mm when measuring static objects of similar dimensions. In order to sample range at video-rates, the images of the scene to be digitised are stored on video tape. This allows the image processing to be performed off-line so the sampling rate of the system only depends on the frame-rate of the video equipment used. The work was motivated by the need to acquire information on mouth shape for acoustics of speech experiments so that the results presented are for measurements of mouth shape and objects relating to the field. The reasons for producing a new system, and the choice of CESL are discussed. The work covers the entire implementation of the range-finder, including code and slide design, feature extraction, feature interpretation, calibration and 3D reconstruction, and performance evaluation. A modification on the Blackman and Tukey classical power density spectrum (PDS) estimator was used for feature extraction which was shown to perform better than other techniques evaluated. The accuracy of detected features and the probability that they were not spurious was determined, based on the feature confidence output by the PDS estimator. Detected features were tracked to produce segments, and encoded as a directed acyclic graph (DAG) which was then matched with the original code sequence using a fast but sub-optimal technique. An existing camera calibration technique was used which was extended to include a second step for projector calibration. The projector model used was linear, had the correct number of degrees of freedom, and was particular to our stripe system.

An automatic data acquisition system for 3D reconstruction in echocardiography

Abstract: Presents a system that allows the acquisition of 2D echographical images that can later be used for the reconstruction of a 3D representation of the human heart. The acquisition process is automatic, so that no interaction by the operator is necessary. The system is based on the so-called 'rotational method' and is conceived such that it can be used with virtually any existing echograph. The 2D images are provided in a digitized form and can be used instantly for further processing. >

Computer-Assisted Analysis and 3D Visualization of Blood Vessels Based on Cone-Beam CT Images

Abstract: In this paper, we present a method for analysis and 3D visualization of high-resolution 3D blood vessels images obtained by cone-beam CT. We are working toward an interactive system which exploits the recent 3D reconstruction approaches, computer graphics, and 3D image processing techniques to facilitate detailed anatomical measurement and visualize the complex blood vessels morphology. The key approaches of our system are a 3D reconstruction image from cone-beam projections, a quantitative analysis based on the blood vessels structure description, and a volume visualization for the inspection of blood vessels structures. From results of the application to a patient's blood vessels, we present the effectiveness of our system.