A novel high-speed 3D imaging system, base don projecting a speckle pattern onto an object and imaging the resulting pattern from multiple angles, is proposed that uses a unique single lens camera system with a rotating off-axis aperture. The rotating aperture allows adjustable non-equilateral spacing between defocused images to achieve greater depth of field and higher sub-pixel displacement accuracy. High resolution ultra fast processing is achieved by recursively correlating image pairs down to diffraction limited image size of the optics. Correlation errors are eliminated during processing by a novel technique based on the multiplication of correlation table elements from one or more adjacent regions. The Gaussian nature of the projected speckle pattern is used to reveal image disparity to sub-pixel accuracy. Processing is accomplished by compressing the images into sparse array format before they are correlated. This system is ideally suited for hardware implementation and circumvents many of the inherent limitations of multi- camera system using FFT spectral based correlation.

Gathering depth data using structured light has been a procedure for many different environments and uses. Many of these system are utilized instead of laser line scanning because of their quickness. However, to utilize depth extraction for some applications, in our case laparoscopic surgery, the depth extraction must be in real time. We have developed an apparatus that speeds up the raw image display and grabbing in structured light depth extraction from 30 frames per second to 60 and 180 frames per second. This results in an updated depth and texture map of about 15 times per second versus about 3. This increased update rate allows for real time depth extraction for use in augmented medical/surgical applications. Our miniature, fist-sized projector utilizes an internal ferro-reflective LCD display that is illuminated with cold light from a flex light pipe. The miniature projector, attachable to a laparoscope, displays inverted pairs of structured light into the body where these images are then viewed by a high-speed camera set slightly off axis from the projector that grabs images synchronously. The images from the camera are ported to a graphics-processing card where six frames are worked on simultaneously to extract depth and create mapped textures from these images. This information is then sent to the host computer with 3D coordinate information of the projector/camera and the associated textures. The surgeon is then able to view body images in real time from different locations without physically moving the laparoscope imager/projector, thereby, reducing the trauma of moving laparoscopes in the patient.

Previous developments of Biris technology for road inspection application sled to the birth of a very rugged high power 3D sensor using a three-line laser projector. In this hostile environment where shock, vibration, dust and dirt are constantly present, 3D ranging must comply with very severe demands. Resolution, precision and reliability are strictly observed. The new approach keeps the same data density but uses high frame rate image capture and feature extraction. With high speed capture, the number of actual laser projections can be reduced to one, without having to decrease the number of measured profiles per second. This eases the data processing by keeping the same characteristics for the projected line from one acquired profile to the next and over the whole acquisition sequence. The data rate achieved maintains the maximum 10 cm profile spacing for the road surface sampling up to speeds of 72km/h. Range measurements and intensity I, are obtained for each measured point. The field-of-view and depth of field for this sensor are very important characteristics which are adapted for the application. However this solution implies a new sensor design and an optimized image capture geometry. Triangulation and defocusing are used to improve the signal to noise ratio and to provide immunity to false measurements.

Pseudo-random point configurations can be used for many vision tasks, both active and passive. Examples are the projection of a pseudo-random light pattern to perform stereo matching or depth estimation by triangulation or robot navigation. The use of the local arrangement of the random features is attractive in some situations because labelling can be performed more robustly than by proliferating the feature types with other coding means. In this context the use of projective invariants provides either a classification measure or an indexing tool which is insensitive to surface position and camera geometry, which have proven invaluable to curb the complexity of the search by order of magnitudes. So far no work has been done on analyzing how these pseudo-random patterns should be like to make the use of invariants more effective, in particular with respect to discrimination and noise sensitivity. This paper addresses this problem for the common case of 5-point projective invariants. A stochastic approximation strategy is employed that iteratively adjusts the position of the points in the pattern while trying to maximize a spacing measure between the invariants. The result clearly illustrate the benefits of the approach which makes optimized pseudo random patterns of identical features a valid alternative to other forms of pattern coding for 3D capture.

The creation of virtual objects for CAD/CAM and animation systems on the base of data gathered by full-field optical measurement of 3D objects is presented. The absolute coordinates are obtained by combined fringe projection/photogrammetry based system with big measurement volume. The procedures for automatic calibration of the measurement volumes, base don sequential localization of a reference plane with markers are explained. To get information about true 3D object, clouds of points from numerous directions are gathered. The calibration procedure for determination of the parameters necessary for merging clouds of points is presented. The new and fully automatic procedure which process the cloud of measured points into triangular mesh accepted by CAD/CAM and multimedia systems is described. The adaptive process of reducing the number of triangles on the base of second derivative of local curvature of objects' surface is explained. The usefulness of the complete process of measurement and virtual object creation is proved by closing the reverse engineering sequence through producing the copy of an exemplary object. Also the virtual objects created are implemented into computer graphics and animation systems.

One problem with moire topography for 3D surface metrology is the so-called 2(pi) -ambiguity limiting the maximum step height difference between two neighboring samples points to be less than half the equivalent wavelength of moire fringes. To cope with the ambiguity problem, a special scheme of scanning moire technique is proposed by resorting to the frequency domain fringe analysis that is in fact originated from white light scanning interferometry. This new more principle of 3D measurement allows determining the absolute height of the surface without information on absolute fringe orders so that largely stepped surfaces are measured with a great improvement in accuracy.

In order to reconstruct the viewable surface o an object completely, multiple views of the same object have to be used and integrated into a common coordinate system. One of the major problems of the 3D surface reconstruction using a turntable, is the varying resolution in the direction to the camera, due to the varying distance of object points to the rotational axis of the turntable. To guarantee a uniform object resolution, we calculate the next angel dynamically, depending on the entropy of the surface part actually acquired. To minimize the loss of information and to guarantee a uniform surface resolution, we derive a relation between the entropy and the next viewing angle, based on the profile sections acquired in the last two steps of the acquisition.

A novel 3D camera named Axi-vision camera was developed. IT can acquire both color and distance information of objects. An intensity-modulated light illuminates objects and the camera with an ultra-fast shutter captures the light reflected from the scene. The distance information is obtained from two images of the same scene taken under linearly increasing and decreasing illuminations. The camera does not require scanning, multiple camera units, or complicated computations. It is possible to acquire distance information for each pixel of a TV image at real time. The operational features and technical specifications of the camera were investigated. The application to TV program production, such as replacing the image of an object at a particular distance by another, was demonstrated. A new 3D display system is also proposed, developed and demonstrated.

In this paper, we propose a fast SFF method, which also provides accurate shape estimation for objects with complex geometry. Dynamic programming is a mathematical tool to determine the optical solution to an n-variable problem efficiently. In the new SF method, dynamic programming is modified and used to find the optimal path that gives maximum focus measure at each pixel with certain constraints. Then the shape of object can be estimated over the 3D FIS. An automated CCD camera system has been developed to implement the proposed method.

This paper addresses a new scheme of acquisition of 3D image representation from range data and texture data. The concept of a layered structure defined for painting, such as long- range, mid-range, and short-range views, that can be applied to a 3D image. Long and mid-range views are located at a reasonable distance, and therefore do not require the perfect 3D structure. Instead of describing the perfect 3D structure, an image can be represented more simply with range information through a plane-model approximation. Setting representation can be used like stage setting to approximate objects and describe a 3D structure by plane- model. It is effective as a simplified means of describing 3D image space as in long and mid-range views that does not require a detailed 3D structure. We can obtain the parameters of plane shape from the integration of range data measured by a 3D laser scanner and a number of still images captured by a digital camera.

This paper in on direct estimation of a dense map of 3D motion and depth from a sequence of images. The proposed method is a gradient-based method which uses the minimum description length principle to formulate the problem as a minimization of a length of description cost function. The method is direct as it does not need priori estimation of optical flow. The minimization is carried out by continuation and resolved numerically by Jacobi iterations. Experimental results are provided.

The methodologies and tools used recently in animation are presented and their most significant problems connected with combining real and virtual world are recognized. It includes creating of computer graphics libraries of realistic 3D objects and describing the models of animation in 3D space. The presented measurement methodology simplifies the process of generation of virtual objects and gives as the result: shape and movement description of the monitored object. Optical structured light methods are proposed for gathering the information about a shape, deformations of the shape and shifts of 3D objects. Authors apply the spatio-temporal approach, in which the spatial analysis of fringe pattern delivers information about initial shape of the object, while the temporal analysis of intensity variation I(t) in the given pixel provides information about out-of-plane displacement. I(t) is analyzed alternatively by adaptive sinusoidal fitting algorithm or by Fourier transform based methods. The comparison between these methods is given and exemplary measurements are presented.

Present work is focused to real time algorithms for estimating 3D structure and motion from an unstructured environment, we assume an uncalibrated camera. Our approach can be divided into two stages. The first is a self- calibration process, which is performed by means of a rotating camera, subsequently a reconstruction process is accomplished by translation movement. In both stages no correspondences are found in an explicit way, alternatively correspondences and transformations matrices are calculated together. To this end, we combine optical flow constraint with a projective model of amoving camera. Additionally we implement a multi-resolution approach to solve the infinitesimal nature of optical flow constraint. We also found this work in the bottleneck time requirements existing in artificial vision systems that needs blending complex algorithms and huge amount of data. Results are presented to show the validity of the developed process.

Retrieving spatial characteristics of subjects form video sequences is a critical issue for many applications. Actually in many cases video streams are the only kind of data it is possible to get on amoving object. In this paper, we present a method for spatial information extraction of amoving subject using multiple video sequences. This method is based on a prediction/correction approach. Spatial characteristics values are converted into synthetic images, and a similarity measure between images is used to estimate their validity. An optimization algorithm is used to adjust the characteristics values, until the best adequation between real and synthetics images is reached. Both the image comparison and the optimization algorithm are designed to handle some problems of this domain: occlusion robustness and noise in images. A set of experiments illustrates the efficiency and the robustness of the approach. Both artificial and real video sequences are used, illustrating various situations.

The objective of this work is to model 3D objects in multiple-object scenes. Towards this objective, we employ deformable simplex meshes first developed by Delingette. These meshes have several convenient characteristics, including the fact that their deformation is controlled with simple geometric entities and is, therefore, computationally inexpensive. We begin with a cloud of 3D points that represents one or more object.s The cloud is then encapsulated with an initial simplex mesh to be subsequently deformed. During deformation, a breaking process enables the mesh to split into multiple meshes, one for each object. The breaking process is based on a 2D net approach introduced by Yoshino and laser refined by Bro-Nielsen. We model the arcs in the mesh as springs that break when their tension exceeds a threshold. This breaking process is additionally controlled by heuristics to prevent breaking on edges of the same object. Once the modeling process is complete, each object is separated from other objects and modeled with its own simplex mesh.

A new method is proposed for building a polyhedral mesh form a set of planar contours representing coordinate constant slices of an object. This method can be used for extracting a boundary surface from volumetric data, a typical application of which is surface reconstruction for objects imaged by CT. With a computational complexity and processing time comparable to those of the Marching Cubes algorithm, an alternative method for extracting iso-surfaces from CT images, this method creates a surface with half of the number of triangles required by Marching Cubes without sacrificing for the accuracy of the surface. This algorithm begins with identifying boundary vertices in three sets of mutually perpendicular planes through the volumetric image data. Using the neighborhood relationship between the vertices in each plane the 3D neighborhood structure of the surface is determined, and either 4 or 6 neighbors are assigned to each vertex. The surface is then tiled with triangles by joining each vertex to two of its immediate neighbors. We applied this method to both computer generated phantom data, and real MR images of human head and compared the result with those of the Marching Cubes algorithm. The experiments show that the proposed method creates more regular and uniform triangulation with smaller number of triangles.

An algorithm for the automatic reconstruction of triangular mesh surface model form range images is presented. The optimal piecewise linear surface approximation problem is defined as: given a set S of points uniformly sampled from a vibrate function f(x,y) on a rectangular grid of dimension W X H, find a minimum triangular mesh approximating the surface with vertices anchored at a subset S' of S, such that the deviation at any sample point is within a given bound of (epsilon) > 0. The algorithm deploys a multi- agent resource planning approach to achieve adaptive, accurate and concise piecewise linear approximation using the L-(infinity) norm. The resulting manifold triangular mesh can be directly used as 3D rendering model for visualization with controllable and guaranteed quality. Due to this dual optimality, the algorithm achieves both storage efficiency and visual quality. The error control scheme further facilitates the construction of models in multiple levels of details, which is desirable in animation and virtual reality moving scenes. Experiments with various benchmark range images form smooth functional surfaces to satellite terrain images yield succinct, accurate and visually pleasant triangular meshes. Furthermore, the independence and multiplicity of agents suggest a natural parallelism for triangulation computation, which provides a promising solution for the real-time exploration of large data sets.

This paper presents a segmentation algorithm for 3D whole body surface scan data. The algorithm is based on 2D projection of 3D data and has achieved good result in a number of limited surface shapes. The method has been successfully employed to extract the torso, arm, and leg segments of the human body.

This paper presents a new acquisition method and the application of the technology for the acquisition of a person's foot. The sensor is designed to meet the requirements of measuring a variety of feet under various ambient optical conditions that can, in other circumstances, seriously affect the data measurements and reduce the reliability of the system. The most important distinction between this ranging method and other more classical approaches is the high tolerance of the method to external optical perturbations as well as reflections form other sensor heads. This allows its use in conventional 'store' or 'medical offices' that are usually bright places.

In this paper we present a new approach for apparel modeling based on voxel specifications that characterize control vertices and patch specifications. Our specification language is geared for robust apparel modeling by enforcing a strict control vertex coding via a combination of a static cross sectional slice and dynamic control polyhedron checking. Unlike most previous approach for apparel are not hard coded into the system. Instead we simply add suitable type definitions to the specification and define patterns to these types. We compile these specifications into a high performance volumetric apparel design system. Important components of our approach include efficient algorithms, for extraction of control vertices form slices of the volume and transformation of these vertices on sequences of such slices. Our system has been tested with the visible female system and we show a couple of examples generated using our approach.

A standard template library has been developed for animation of voxel humans. The template class helps to standardize natural animations of humans, embed the data structures in containers, and also enables reuse of associated algorithms. The validity of this new approach and its utility is demonstrated by experimental results

We describe a project to construct a 3D numerical model of Michelangelo's Florentine Pieta to be used in a study of the sculpture. Here we focus on the registration of the range images used to construct the model. The major challenge was the range of length scales involved. A resolution of 1 mm or less required for the 2.25 m tall piece. To achieve this resolution, we could only acquire an area of 20 by 20 cm per scan. A total of approximately 700 images were required. Ideally, a tracker would be attached to the scanner to record position and pose. The use of a tracker was not possible in the field. Instead, we used a crude-to-fine approach to registering the meshes to one another. The crudest level consisted of pairwise manual registration, aided by texture maps containing laser dots that were projected onto the sculpture. This crude alignment was refined by an automatic registration of laser dot centers. In this phase, we found that consistency constraints on dot matches were essential to obtaining accurate results. The laser dot alignment was refined by an automatic registration of laser dot centers. In this phase, we found that consistency constraints on dot matches were essential to obtaining accurate results. The laser dot alignment was further refined using a variation of the ICP algorithm developed by Besl and McKay. In the application of ICP to global registration, we developed a method to avoid one class of local minima by finding a set of points, rather than the single point, that matches each candidate point.

Image-based reconstruction from randomly scattered views is a challenging problem. We present a new algorithm that extends Seitz and Dyer's Voxel Coloring algorithm for reconstructing a voxelized representation of 3D object from a series of images. Voxel Coloring traverses a discretized 3D space in 'depth order' to identify voxels that have a unique coloring, constant across all possible interpretation of the scene. This approach has several advantages over existing stereo and structure-from-motion approaches to scene reconstruction. First, the technique can handle a great magnitude of visibility change. So, the cameras can be positioned far apart without degrading accuracy or run-time. Second, the technique integrates numerous imags to yield dense reconstruction without degrading run-time. Unlike Seitz and Dyer's algorithm, ours considers the perspective projection effect on the voxel size. We also present a different search method that traverses the voxels along the projection rays of the images. The new search method optimizes the search process by employing the geometry constraint of the pinhole camera model. In the new search method, the size of the voxels is non uniform depending on the distance from the images. Experimental results for simulated and real image sequences show the efficiency of our algorithm.

Modeling of 3D objects from image sequences is a challenging problem and has been a research topic for many years. Important theoretical and algorithmic results were achieved that allow to extract even complex 3D models of scenes from sequences of images. One recent effort has been to reduce the amount of calibration and to avoid restrictions on the camera motion. In this contribution an approach is described which achieves this gaol by coming state-of-the-art algorithms for uncalibrated projective reconstruction, self- calibration and dense correspondence matching.

Restoration of a broken pottery for repair and reconstruction is the main aim of this project. IRAN is a country with more than 2500 years civilization. Archaeologists dug out potters which most of them were broken or in fragments. These potteries are our heritage and track tour ancient civilization; therefore, reconstruction of them is very important. The project is searching to find a way for restoration of the broken earthenware in the computer by using close range photogrammetry technique and CAD system. An earthenware was photographed and its model was restored in the computer by using a P33 stereo plotter and MicroStation. Then the earthenware was broken to small pieces and these parts were photographed again. Models of small pieces constructed in computer. Finally, a method was developed to restore the broken earthenware on the computer according to the initialized model.

Towards photo-realistic 3D scene reconstruction form range and color images, we present a statistical technique for multi-modal image registration. Statistical tools are employed to measure the dependence of tow imags, considered as random distributions of pixels, and to find the pose of one imaging system relative to the other. The similarity metrics used in our automatic registration algorithm are based on the chi-squared measure of dependence, which is presented as an alternative to the standard mutual information criterion. These two criteria belong to the class of information-theoretic similarity measures that quantify the dependence in terms of information provided by one image about the other. This approach requires the use of a robust optimization scheme for the maximization of the similarity measure. To achieve accurate reslut, we investigated the use of heuristics such as genetic algorithms. The retrieved pose parameters are used to generate a texture map from the color image, and the occluded areas in this image are determined and labeled. Finally the 3D scene is rendered as a triangular mesh with texture.

Traditional methods of linear based imaging limits the viewer to a single fixed-point perspective. By means of a single lens multiple perspective mirror system, a 360-degree representation of the area around the camera is reconstructed. This reconstruction is used overcome the limitations of a traditional camera by providing the viewer with many different perspectives. By constructing the mirror into a hemispherical surface with multiple focal lengths at various diameters on the mirror, and by placing a parabolic mirror overhead, a stereoscopic image can be extracted from the image captured by a high-resolution camera placed beneath the mirror. Image extraction and correction is made by computer processing of the image obtained by camera; the image present up to five distinguishable different viewpoints that a computer can extrapolate pseudo- perspective data from. Geometric and depth for field can be extrapolated via comparison and isolation of objects within a virtual scene post processed by the computer. Combining data with scene rendering software provides the viewer with the ability to choose a desired viewing position, multiple dynamic perspectives, and virtually constructed perspectives based on minimal existing data. An examination into the workings of the mirror relay system is provided, including possible image extrapolation and correctional methods. Generation of data and virtual interpolated and constructed data is also mentioned.

We propose a new ray-space interpolation scheme using an adaptive filter. Unlike the previous works related to view interpolation, which detect disparities, our scheme adopts a simple signal processing method; adaptive filtering to Epipolar Plane Image (EPI). First, the original EPI is up- sampled. Then, for each pixel to be interpolated, the block surrounding the pixel is analyzed and the best filter is selected according to the analysis. The filter set includes various interpolation filters with different directionality. Finally, we apply the filter to up-sampled EPI and generate the intermediate ray-space data. Since our scheme does not need pattern matching, it requires less computation cost than the conventional interpolation schemes, and therefore, it is very fast and suitable for hardware implementation. Experimental results show that the proposed scheme interpolates ray-space data with higher PSNR than other interpolation methods, such as nearest neighbor interpolation, the linear interpolation and block matching interpolation.

In this paper, we investigate the impact of intensity edge maps (IEMs) on the segmentation of noisy range images. Two edge-based segmentation algorithms are considered. The first is a watershed-based segmentation technique and the other is the scan-line grouping technique. Each of these algorithms is implemented in two different forms. In the first form, an IEM is fused with the range edge map prior to segmentation. In the second form, the range edge map alone is used. The performance of each algorithm, with and without the use of the IEM information, is evalute and reported in terms of correct segmentation rate. For our experiments, two sets of real range images are used. The first set comprises inherently noisy images. The other set is compared of images with varying levels of artificial, additive Gaussian noise. The experimental results indicate that the use of IEMs can significantly improve edge-based segmentation of noisy range images. Considering these result, it seems that segmentation tasks invovling range images captured by noisy scanners would benefit from the use of IEM information. Additionally, the experiments indicate that higher quality edge information can be obtained by fusing range and intensity edge information.

The fast, contact-free and highly precise shape measurement of technical objects is of key importance in the scientific- technological area as well as the area of practical measurement technology. The application areas of contact- free surface measurement extend across widely different areas, e.g., the automation of production processes, the measurement and inspection of components in microsystem technology or the fast 3D in-vivo measurement of human skin surfaces in cosmetics and medical technology. This paper describes methodological and technological possibilities as well as measurement technology applications for fast optical 3D shape measurements using micromirror-based high-velocity stripe projection. Depending on the available projector and camera facilities, it will be possible to shoot and evaluate compete 3D surface profiles within only a few milliseconds.