Showing papers on "3D reconstruction published in 1992"

3D computer vision and applications

Abstract: In order to overcome the shortcoming of the current stereo vision, a reliable stereo method is proposed making use of the reliability of features in determining correspondence. A rich description of a scene is obtained by integrating range data and color data. Interpretation of a scene requires the use of the knowledge or models of objects. Based on the knowledge, most effective process is triggered at each stage depending on what kind of information is obtained at that stage. In this paradigm, a monocular image of a 3D scene can also be interpreted. The author describes applications of 3D computer vision and his approach to a more flexible 3D computer vision. >

Active photometric stereo

Abstract: An active vision technique for determining the absolute depth of surfaces is described. The algorithm assumes a very general model for the reflectance properties of the surface, and is valid for most of the shading models commonly used in computer vision work. The algorithm relies on the controlled motion of a point light source, which is not at infinity but relatively close to the surface and to the camera. The sensitivity of the computed depth values to errors in the measured quantities is derived, allowing a confidence measure for the depth to be determined. The confidence measure can aid in the estimation of accurate depth values from multiple image measurements taken over time. A method based on robust estimation that permits an unbiased estimate of the depth values to be obtained is presented. The results of experiments on synthetic and real-world imagery are reported, illustrating the efficacy of the active photometric stereo algorithm. >

Evaluation of a 3D reconstruction algorithm for multi-slice PET scanners

Abstract: A fully 3D reconstruction algorithm based on filtered backprojection was evaluated for the reconstruction of data obtained with multi-slice positron emission tomography (PET) scanners which have had the septa removed. This algorithm uses forward-projection through the reconstructed images of a 2D subset of the data to complete the 3D dataset thus satisfying the condition of shift invariance. This is followed by 3D filtered backprojection. Axial sampling was doubled by combining adjacent polar angles, thus improving reconstructed axial resolution. The algorithm was tested using real and simulated datasets and gave high quality reconstructions without artifacts over a wide range of imaging conditions. Events are placed accurately throughout the imaging volume as determined by measurements with a MRI/PET registration phantom. The forward-projection step leads to degradation in image resolution due to insufficient axial and transaxial sampling. This effect is amplified if multiple iterations of the algorithm are used, with little decrease in image noise. Changing the filter employed in the initial 2D reconstruction can be used to alter the noise and resolution characteristics of the 3D images.

3D-surface reconstruction of intravascular ultrasound images using personal computer hardware and a motorized catheter control.

Abstract: A system for three-dimensional (3D) presentation of intravascular ultrasound (US) images was designed. A standard hardware configuration based on personal computer equipment was used for acquisition and processing of image data. Pullback imaging with the US catheter was controlled by a specially designed motor assembly and performed either in equidistant 1–10 mm steps or at a constant retraction speed. Curvature of the vessel was documented on biplane digital subtraction angiograms and the US images were arranged according to the vessel shape in the angiograms. 3D reconstruction appears essential for spatial orientation of intravascular US and may be helpful in the planning of vascular interventions for a better appreciation of the extent and morphology of vascular lesions.

Critical Sets for 3D Reconstruction Using Lines

Abstract: This paper describes the geometrical limitations of algorithms for 3D reconstruction which use corresponding line tokens. In addition to announcing a description of the general critical set, we analyse the configurations defeating the Liu-Huang algorithm and study the relations between these sets.

A knowledge based system for diagnosing spinal deformations: moire pattern analysis and interpretation

Abstract: For young people, spinal deformations are an evolving process which must be detected and treated as early as possible. The moire technique is quite simple, not expensive, not aggressive and well suited to the detection of spinal rotations. Therefore, a computer vision system for recognition of spinal deformations would be very useful for decision support. The authors present a knowledge-based system for diagnosing spinal deformations from human back moire image. To realize this computer system, the approach consists of extracting fringe contours from moire images using the infinite size symmetric exponential filter which is an optimal smoothing filter, then localizing some relevant elements such as the centers of shoulder-blades and the spinal column to make measurement which are crucial for the diagnosis process, and finally performing 3D reconstruction and diagnosis. >

Structure from motion using an active vision paradigm

Abstract: A method for the reconstruction and localization of geometrical primitives using active dynamic vision is presented. The approach is based on the use of the interaction matrix related to the visual data describing a primitive. Next, active vision is considered by computing adequate camera motions with a control law in closed-loop with respect to visual data. Simulation results on the localization of a sphere are presented and show that active vision can to a large extent improve the accuracy of the structure estimation. >

Real-time stereo vision with multiple arrayed camera

Abstract: Real-time stereo vision using a multiple arrayed camera system, named MAC vision, is proposed. MAC vision processes the correspondence problem, one of the basic problems in stereo vision, very rapidly by making use of the simple geometric relationship of images on the same-height scanning lines. Several problems involved in MAC vision are discussed. The results of experiments performed on a newly constructed five-camera prototype are presented. The experimental results showed that a speed of 16 (frames/s), where each frame included 1800 points. >

3D-pose estimation from a quadratic-curved feature in two perspective views

Abstract: Estimation of 3D information from 2D-image coordinates is a fundamental problem in both machine vision and computer vision. A closed-form analytical solution is presented to the problem of 3D-location estimation of quadratic-curved features based on two images. Compared to previous solution methods, the proposed method has the advantages that it is a closed-form solution method, it is mathematically simpler, and a priori knowledge of the size and shape of the feature is not required. The method is demonstrated for an elliptical feature, using a simulated experimental set-up. >

3D Motion and reconstruction of coronary networks

Abstract: This paper describes an original method combining spatial and temporal features for the 3D reconstruction of coronary arteries from two X-ray views. Motion estimation is performed along the vessel centerlines in the two images by means of an optical flow based method. Centerline tracking and matching is carried out through epipolar constraints. This scheme allows to process automatically these two overall image sequences. Its effectiveness has been demonstrated on real data and the results are discussed in terms of expected benefits and potential limitations.

A method for the 3D reconstruction of indoor scenes from monocular images

Abstract: The recovery of the 3D structure of indoor scenes from a single image is an important goal of machine vision Therefore, a simple and reliable solution to this problem will have a great influence on many tasks in robotics, such as the autonomous navigation of a mobile vehicle in indoor environments.

Exploratory active vision: theory

Abstract: An active approach to the integration of shape from x modules-here shape from shading and shape from texture-is proposed. The question of what constitutes a good motion for the active observer is addressed. Generally, the role of the visual system is to provide depth information to an autonomous robot; a trajectory module will then interpret it to determine a motion for the robot, which in turn will affect the visual information received. It is suggested that the motion can also be chosen so as to improve the performance of the visual system. >

Towards 3D vision from range images: an optimization framework and parallel distributed networks

Abstract: This thesis investigates approaches to object recognition in computer vision. The starting point of the investigation is depth, or range, images of visible object surfaces. At the representation level, we use attributed relational graphs (ARGs), for object-centered descriptions of surfaces in images as well as models. An ARG encodes invariant properties such as curvature of surfaces, and relations such as distances and normal angles between them. At the computational level, the main problems studied concern firstly the extraction of ARG representation from range images, which involves the issues of continuity-controlled smoothing and reliable computation of surface curvature, and secondly inexact matching between the image ARG and object model ARGs. First we propose a unified computational framework for solving low, intermediate and high level computer vision problems in 3D object recognition from range images. All three levels of computation are cast in an optimisation framework and are suited for implementation on parallel distributed, or neuron-like, architectures. In the low level computation, the tasks are to estimate the curvature images from the input range data, in which depth discontinuities are taken into account to avoid oversmoothing. Subsequent processing at the intermediate level is concerned with segmenting these curvature images into coherent curvature sign maps. At the high level, image features are matched against model features based on the object-centered ARG representation. It is shown that the above computational tasks at each of the three different levels can all be formulated as optimising an energy function. The optimisation is performed using parallel and distributed relaxation-based algorithms which are well suited for neural network implementation. We then further develop the low level work by proposing a class of adaptive regularizer (AR) for continuity-controlled smoothing. We analyze mechanisms by which the AR works. The analysis shows that the fundamental difference between different regularizes lies in properties of interaction between neighboring points, which relates to the smoothing strength, determined by these regularizes. Based on this, we derive conditions under which regularizes are adaptive to discontinuities to avoid oversmoothing. The analysis also shows that adaptive regularization solutions have a closer affinity to solving Euler equations than to minimising some close-formed energy functionals. We further develop the high level work by proposing a parallel distributed approach for matching and recognizing overlapping partially occluded objects. Simultaneous separation and inexact matching of superimposed sub-ARGs is performed by optimising a global gain functional using relaxation labelling methods. We suggest that the approach could potentially perform simultaneous recognition of overlapping free-formed 3D objects directly from range data without segmentation: Segmentation, matching and recognition are all done once the direct mapping from pixels to models is found. The algorithm is inherently parallel and distributive and could be efficiently implemented on commercially available SIMD architectures such as the Connectionist Machine.

3D reconstruction of small underwater objects using high-resolution sonar data

Abstract: A high-resolution, very compact acoustic lens sonar was used to investigate how well small underwater objects can be reconstructed in 3D. The purpose was to develop an acoustic snapshot approaching photographic quality in turbid water where cameras and lens are ineffective. In-water experiments have provided data from an 8-by-8 array of acoustical beams with 1.5 degree conical resolution. Algorithms have been developed to interpolate the conical grid onto a grid. Once gridded, a volume rendering algorithm was applied to produce a full 3D rendering of the object. The sensitivity of this process to different gridding algorithms is evaluated. It is shown that objects can be quantified in size quite precisely. The degree to which shape can be extracted and small details made visible is discussed. The authors also consider how well the reconstruction can be improved by using multiple passes over the object. Registration algorithms are presented, and it is shown that the alignment process can be done with enough accuracy to improve the image and provide greater detail about the object. >

Determining when 3D reconstruction methods are required in volume PET imaging

Abstract: An attempt is made to determine at which point a 3-D reconstruction algorithm is needed for a volume imaging scanner as the axial extent is increased. From simulations of an existing volume PET (positron emission tomography) scanner with a maximum axial acceptance angle ( phi ) of 9 deg, it is shown that a two-dimensional image reconstruction with alpha =9 deg is a compromise between sensitivity, speed, and accuracy compared to a two-dimensional reconstruction with a software limit of alpha =1 deg and a three-dimensional reconstruction with alpha =9 deg. It is also shown that a proposed scanner with alpha =27 deg requires three-dimensional reconstruction in order to achieve maximum sensitivity without unacceptable losses in accuracy. >

Photometric models in multispectral machine vision

Abstract: The performance of several tasks in multispectral computer vision involves assumptions about the reflection of light from surfaces. These tasks include color constancy (visual representation of spectral reflectances independent of the illuminant spectrum), object-based image segmentation, and deduction of the shape of a surface from its shading. Most color-constancy theories implicitly assume Lambertian, coplanar reflecting surfaces, a distant viewer, and a distant light source that may have many components that are spatially and spectrally distinct. Object-based-segmentation theories allow curved surfaces, each of whose scattering kernels is the sum of a few separable terms (each of which is the product of a wavelength-dependent part and a geometry-dependent part). There is no restriction on the distances of light sources or observer. However, for these theories the illuminant angular/spectral distribution must consist of only one or two separable terms. Finally, A. Petrov's shape-from-shading theory allows the light source to have nearly arbitrary spectral and spatial composition, but requires the surface scattering kernels to have Lambertian dependence on the surface normal. The present paper compares these photometric models.© (1991) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

Fusion of stereo and motion vision for 3-D reconstruction

Abstract: Describes a robotic vision system based on the fusion of stereo and motion vision for 3-D reconstruction. Its performance under occlusions is improved by considering the image matching noise dynamics in the motion module. The stereo module is based on the maximum a posteriori estimate of depth given the measured stereo disparity. Simulations with noisy synthetic images showed that the system is capable of 3-D reconstruction and of resolving occluding from occluded objects with potential for real-time operation. >

Visual reconstruction with adaptive and arbitrarily oriented meshes

Abstract: Surface reconstruction and visualization from sparse and incomplete surface data is a fundamental problem and has received growing attention in both computer vision and graphics. The paper introduces a novel approach to recovering and visualizing complex shapes from range data, in which a surface of unknown shape is reconstructed with view-invariant characteristics on arbitrary viewplanes from sparsely sampled surface range data so that the entire surface can be visualized from any view points with constant. The authors develop a graphical reconstruction model based on differential geometry using adaptive and arbitrarily oriented meshes, in which the number of surface nodes is generated dynamically on an active viewing plane and then increased in number and influence one another until the entire surface has been interpolated with a sufficient accuracy. The computation results on real range data of human faces are shown. >

Active vision with two differentiated visual fields

Abstract: A vision system is proposed that has two differentiated visual fields, i.e. peripheral and central, and the fixation point to the scene is actuated actively. The proposed vision system is aiming at an active vision system that recognizes environment actively by gazing at each object of interest. To build this system, the authors developed a vision system with two CCD cameras of which fixation points are controlled by a computer. A fundamental vision model is proposed that consists of five independent neural network modules. Each module is educated individually and processes visual information independently. To signify the feasibility of the proposed vision model, some fundamental experiments were carried out. >

Computer-assisted 3D-reconstruction and statistics of the limbic system. 1. Computer-assisted 3D-reconstruction of the hippocampal formation, the fornix, and the mamillary bodies.

Abstract: The hippocampal formation of eight perfusion-fixed human brains was examined using new methods according to stereotactic and morphometric principles (macrovibratome and computer-aided 3D reconstruction). The reconstructions form part of a neuroanatomical reference system (NeuRef). This reference system allows for 3D visualisation of the brain and its components on a computer graphic workstation, as well as for the presentation of the union set based on a neuroanatomical structure taken from this sample of brains. This retrievable knowledge of neurofunctional systems is important for the preoperative planning of neurosurgeons and the adjustment of radiotherapy.

3d Reconstruction of Plants.

Abstract: Tlv description of ~ccrrilctrica~ slructurc o f vcpctation canapics in prncrnl. rnrl i r l pnr~iculer, o r crops has today scvcral npplicarinr~s : 3 Crop s i~ i i t ~ la t i~ ig lilntlelc necrl to enter, ngrollornicnl vnriat~lcs. at a givcn (late. SIIC~ as lcnf arcas, op~icnl propcrlics. ... 2 Tlic clinrnc~cri~.atinn of crnp srnlclurc using non ~Ic

Control in the bootstrap phase of a computer vision system

Abstract: A sensor control algorithm is developed which takes into account errors in the position of objects stored in the long term memory of the vision system, errors in image measurement and uncertainties in object attributes. It is tested on synthetic and real data to demonstrate its validity.

Stochastic structure estimation by motion

Abstract: A field of great interest in computer vision is depth reconstruction by motion. The final goal is the computation of the visible surface structure in a 3D scene by analyzing a sequence of digital images acquired moving a camera in the environment. This paper describes a method of depth reconstruction based on stochastic modeling of the motion, the image acquisition processes, and the 3D-2D projection. The stochastic model is based on the well-known extended Kalman filter to derive an optimized depth estimation: it integrates successive views by using a pair of optical flow equations that we have adapted to a general pin-hole camera model (linear transformation from 3D to 2D coordinates). In comparison with similar methods we developed a reconstruction system to improve the speed of the estimation process and its stability by means of a multi-scale approach and used a massive parallel MIMD machine to speed up globally the estimation process.© (1992) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

Extraction of surface orientation using gray level difference statistics

Abstract: The Processes to reconstruct a 3D shape from a 2D image is one of the important problems in computer vision. In this paper we deal with the problem to extract the object surface orientation from a monocular view image, which is necessary in 3D reconstruction. Generally the process becomes ill-posed problem, because the 3D shape of an object is condensed onto the image by the projection. Therefore the solution of the orientation is not guaranteed to be unique, unless some supplement information is introduced about the object or the surface.