Range Image

Tree skeleton computation is of significance in the geometric modeling of botanic trees and in the application of forestry. This paper describes an approach to extract branch skeletons from a single range image of a tree. A basis of this approach is that the trunk and branches are modeled by generalized circular cylinders, and the tree skeleton is defined as the connected curve-axes of these cylinders. In the proposed system, the range image is partitioned into patches at first based on the discontinuity of depth and axis direction, where each patch only contains points from the same branch. Then each patch is fitted with a series of circular cylinders. Finally the tree skeleton is generated by sequentially connecting the skeleton points of fitted cylinders. This work shows that cylinder fitting can be used to handle the incompleteness of input data, and generate accurate skeleton points and corresponding radii. The main contribution of this paper is that we proposed a new definition and computation of tree skeletons and introduced an efficient and robust cylinder fitting method. Experiment shows the effectiveness of this approach.

Range images are vision and robotics ment of fast and acquisition of 3D the other hand, in order tions may be succes!;fully constraints, such as processing. In this 1 focused on the efficient mations from range simplifying further segmentation [4]. This work focuses positions where a range to determine the surfaces with the aim of obtaining as the one described in view problem. As in el Garciat Susana VelBzquez* Science Engineering gaining popularity in the computer dommunities owing to the developlowcost range sensors that allow the information directly from a scene. On that this kind of dense representaused in fields with real time robotics, it is necessary to simplify its ne, previous work by the authors extraction of triangular approxiimages [2][5] with the aim of processing algorithms, such as image on the problem of determining the sensor should be placed in order of the objects present in a scene, a 3D model of that scene, such [3]. This is known as the next-bestseveral previous proposals, it is

This paper presents an eficient algorithm for generating adaptive triangular meshes from dense range images. The proposed technique consists of two stages. First, a quadrilateral mesh is generated from the given range image. The points of this mesh adapt to the surface shapes represented in the range image by grouping in areas of high curvature and dispersing in low-variation regions. The second stage splits each quadrilateral cell obtained before into two triangles. Between the two possible pips, it is chosen the one whose diagonal's direction is closest to the orientation of the discontinuities present in that cell. Both stages avoid costly iterative optimization techniques. Results with real range images are presented. They show low CPU times and accurate triangular approximations of the given images.

This paper presents a technique for lossy compression of dense range images. Two separate compression schemes are applied. The first scheme (geometric compression) reduces redundant geometric information by generating an adaptive 3D triangular mesh that approximates the shapes present in the original range image. Geometric compression is used for obtaining an efficient representation of the range image that allows further processing. The second compression scheme (topological compression) encodes the connectivity information contained in the triangular mesh. Topological compression is used for generating a compact representation suitable to be stored or transmitted. Both compression schemes avoid costly iterative optimization algorithms. Results with real range images are presented.

A new technique for approximating range images with adaptive triangular meshes ensuring a user-defined approximation error is presented. This technique is based on an efficient coarseto-fine refinement algorithm that avoids iterative optimization stages. The algorithm first maps the pixels of the given range image to 3D points defined in a curvature space. Those points are then tetrahedralized with a 3D Delaunay algorithm. Finally, an iterative process starts digging up the convex hull of the obtained tetrahedralization, progressively removing the triangles that do not fulfill the specified approximation error. This error is assessed in the original 3D space. The introduction of the aforementioned curvature space makes it possible for both convex and nonconvex object surfaces to be approximated with adaptive triangular meshes, improving thus the behavior of previous coarse-to-fine sculpturing techniques. The proposed technique is evaluated on real range images and compared to two simplification techniques that also ensure a user-defined approximation error: a fine-to-coarse approximation algorithm based on iterative optimization (Jade) and an optimization-free, fine-tocoarse algorithm (Simplification Envelopes).

This paper presents an efficient technique for generating adaptive triangular meshes from range images. The algorithm consists of two stages. First, a user-defined number of points is adaptively sampled from the given range image. Those points are chosen by taking into account the surface shapes represented in the range image in such a way that points tend to group in areas of high curvature and to disperse in low-variation regions. This selection process is done through a noniterative, inherently parallel algorithm in order to gain efficiency. Once the image has been subsampled, the second stage applies a two and one half-dimensional Delaunay triangulation to obtain an initial triangular mesh. To favor the preservation of surface and orientation discontinuities (jump and crease edges) present in the original range image, the aforementioned triangular mesh is iteratively modified by applying an efficient edge flipping technique. Results with real range images show accurate triangular approximations of the given range images with low processing times.

A new technique is presented for determining the positions where a range sensor should be located to acquire the surfaces of a complex scene. The algorithm consists of two stages. The first stage applies a voting scheme that considers occlusion edges. Most of the surfaces of the scene are recovered through views computed in that way. Then, the second stage fills up remaining holes through a scheme based on visibility analysis. By leaving the more expensive visibility computations at the end of the exploration process, efficiency is increased.

Recent developments in non-contact industrial range imaging systems have generally focussed either on very highprecision, short range, active triangulation systems capable of detailing measurement volumes of a few square metres or medium precision, time-of-flight systems with typical standoff distances of 2-50 metres. This paper will highlight a range of applications for medium precision (+/-5mm) imaging of relatively complex, room-sized environments which are not currently well served by such systems. It will also suggest that the development of relatively short baseline active triangulation systems may lead to affordable, compact and robust short-range measurement systems capable of delivering range data suitable for such applications. A brief review of current systems for non-contact range-imaging will be followed by a detailed discussion of system trials undertaken with a modified Biris triangulation system in a typical engineering environment containing a wide variety industrial surfaces with a range of surface reflectivity characteristics. The authors will conclude that such a system may offer a compact and flexible alternative to competing technologies that could be readily deployed in complex small-scale industrial settings. Many authors have documented an increasing demand for as-built modelling of large industrial facilities. Applications of such models range from very accurate surveys of tie-in locations to ensure first-time fit of new components to relatively crude volumetric models suitable for planning access or general facility management operations. One of the difficulties associated with such surveys is the development of appropriate specifications to ensure that measurements are captured at an appropriate density for the anticipated application. A number of measurement techniques have emerged to serve such markets. These include conventional geodetic surveying techniques based on total station intersection or radiation, close-range photogrammetry and, increasingly, the application of range-imaging sensors. Theodolite based surveys offer relatively high precision at the expense of comparatively slow data capture times. Since reduction of time spent on site is often of critical importance in such applications these methods are most commonly used for the accurate survey of a few tie-in points required for the fitting of new components and pipe-spools. Photogrammetric surveys can, in certain circumstances, offer comparable measurement precision and have the additional advantage that they enable very rapid data capture and potentially offer complete coverage of the facility for offline measurement. Unfortunately this offline measurement task remains a labour intensive operation that has been stubbornly resistant to attempts to automate image segmentation and object recognition in complex scenes. Against this background range imaging devices appear to have much to offer in that they enable the automatic derivation of dense clouds of 3D on the surfaces of the object. These point clouds offer the potential for some degree of automation of object extraction and thus raise the prospect of cheaper As-Built CAD modelling. Unfortunately many of the early range-imaging systems were often either very expensive, bulky or slow thus offering less flexibility than photogrammetric systems in the very cluttered environments typically encountered in many petrochemical and nuclear environments.

A research on-detecting on-tree mango fruits is presented in this paper. The digital image is used as an input for this research. The method chose is texture analysis and Randomized Hough Transform (RHT). The output for this system is the centre of ellipse which is later defined as location of the mango fruit. This paper tackled the issue of separating overlapping mango fruit and mango fruit that is occluded by leaves. The centre of the ellipse is generated as the output and this information will be used as the grasping point for the fruit picking robot during the picking process. By using this methods and flow, the system successfully detect the fruit including the overlapping one.

To assess if a given manufactured artifact conform to a template CAD model, the artifact is measured by a coordinate measuring system to obtain a point cloud. The point cloud is compared to the CAD model by fitting the point cloud to the CAD model. One of approaches is based on a method required to determine footpoint parameters. The footpoints are the projection of the points in the point cloud onto the CAD surface. In this paper we introduce a new efficient method for estimating footpoint parameters.

This paper describes a 3D scene analysis system that is capable of modelling real-world scenes, based on data acquired by a Laser Range Finder on board of a mobile robot. Laser range images directly provide access to three dimensional information, as opposed to intensity images. This characteristic makes range data suited to build 3D models of objects, and in general real world scenes. Results from the complete modelling procedure are presented, starting with the extraction of surface characteristics from range images, approximation of the surface geometry by a 3D triangular surface mesh and the merging of the surface descriptions from different viewpoints into a consistent 3D scene model. Direct applications of this technique include retrieval of architectural andlor industrial plans into a CAD model, design verification of buildings and plants, and update of a CAD model with the real world ("as-built").

Designing robots for urban search and rescue (USAR) is an extremely challenging task, because it violates many of the assumptions made in the average robotics lab. Robocup Rescue is an international competition that provides an opportunity to measure performance in such environments objectively. The competition allows for the evaluation of new and evolving sensing, hardware and software technologies in an environment that is as close to realistic as possible. This paper discusses the design and implementation of CASTER, the robot that helped Team CASualty come 3rd in the 2005 competition. The strengths of CASTER compared to other teams include high quality victim information, excellent maps and an intuitive user interface. The mapping techniques in particular are novel because they take advantage of a 3D time-offlight range imager, a technology that has only recently become widely available. The results are all the more impressive when one considers that this was the team's first entry into the Robocup Rescue Robot League and that the robot base was delivered seven months before the competition.

This paper describes a 3D scene analysis system that is capable of modelling real-world scenes, based on data acquired by a Laser Range Finder on board of a mobile robot. Laser range images directly provide access to three dimensional information, as opposed to intensity images. This characteristic makes range data suited to build 3D models of objects, and in general real world scenes. Results from the complete modelling procedure are presented, starting with the extraction of surface characteristics from range images, approximation of the surface geometry by a 3D triangular surface mesh and the merging of the surface descriptions from different viewpoints into a consistent 3D scene model. Direct applications of this technique include retrieval of architectural andlor industrial plans into a CAD model, design verification of buildings and plants, and update of a CAD model with the real world ("as-built").

This paper focuses a comparative evaluation of our framework for 3D face recognition and state-of-theart systems. Our method uses a Simulated Annealingbased approach (SA) for range image registration with the Surface Interpenetration Measure (SIM) as the similarity measure, in order to match two face images. The authentication score is obtained by combining the SIM values corresponding to the matching of four different face regions. Experiments were performed on the FRGC v2 database simulating both verification and identification systems and the obtained results were compared to those reported in the literature. By using all the images in the database, a verification rate of 95.9% was achieved, at a False Acceptance Rate (FAR) of 0.1%. In the identification scenario, a rank-one accuracy of 99.5% was obtained. To our knowledge, this is the best rank-one score obtained on the FRGC v2 database, as compared to previously published results.

Figure 1: A range image set and its parametrization. A set of range images U i is captured by direct measurement of real world objects or by means of digital rendering of virtual objects. Each range scan U i is mapped over a domain D i by a parametrization function c i , in a globally consistent way. Among other uses, a semi-regular quad remeshing of the original object can be obtained by regularly sampling ∪D i .

A phased array technique has been mainly utilized for the atmospheric radar and the wind profiler radar, and also applied to the weather radar for research in recent years. As a further development of phased array technology, “Multiple-input multipleoutput (MIMO) technique” , developed in the communication systems, has been applied to the radar. One of the effects is that it is possible to create the virtual antenna aperture plane beyond the actual antenna, and we can also say that it is possible to make the actual antenna size smaller compared to the conventional antenna while keeping the angular resolution. This effect is expected to reduce costs which is the most serious problem to expand phased array radars instead of parabolic antenna systems. In order to confirm the effect, an experimental observation was done using the MU radar and the preliminary result will be introduced. MIMO technique requires orthogonal waveforms on each transmitter to identify the transmitted signals wi...

A research on-detecting on-tree mango fruits is presented in this paper. The digital image is used as an input for this research. The method chose is texture analysis and Randomized Hough Transform (RHT). The output for this system is the centre of ellipse which is later defined as location of the mango fruit. This paper tackled the issue of separating overlapping mango fruit and mango fruit that is occluded by leaves. The centre of the ellipse is generated as the output and this information will be used as the grasping point for the fruit picking robot during the picking process. By using this methods and flow, the system successfully detect the fruit including the overlapping one.

In this paper we propose a new general framework for the application of the nonlinear finite element method (FEM) to nonrigid motion analysis. We construct the models by integrating image data and prior knowledge, using well-established techniques from computer vision, structural mechanics, and computer-aided design (CAD). These techniques guide the process of optimization of mesh models. Linear FEM proved to be a successful physically based modeling tool in solving limited types of nonrigid motion problems. However, linear FEM cannot handle nonlinear materials or large deformations. Application of nonlinear FEM to nonrigid motion analysis has been restricted by difficulties with high computational complexity and noise sensitivity. We tackle the problems associated with nonlinear FEM by changing the parametric description of the object to allow easy automatic control of the model, using physically motivated analysis of the possible displacements to address the worst effects of the noise, applying mesh control strategies, and utilizing multiscale methods. The combination of these methods represents a new systematic approach to a class of nonrigid motion applications for which sufficiently precise and flexible FEM models can be built. The results from the skin elasticity experiments demonstrate the success of the proposed method. The model allows us to objectively detect the differences in elasticity between normal and abnormal skin. Our work demonstrates the possibility of accurate computation of point correspondences and force recovery from range image sequences containing nonrigid objects and large motion.

In this paper we propose a new general framework for the application of the nonlinear finite element method (FEM) to nonrigid motion analysis. We construct the models by integrating image data and prior knowledge, using well-established techniques from computer vision, structural mechanics, and computer-aided design (CAD). These techniques guide the process of optimization of mesh models. Linear FEM proved to be a successful physically based modeling tool in solving limited types of nonrigid motion problems. However, linear FEM cannot handle nonlinear materials or large deformations. Application of nonlinear FEM to nonrigid motion analysis has been restricted by difficulties with high computational complexity and noise sensitivity. We tackle the problems associated with nonlinear FEM by changing the parametric description of the object to allow easy automatic control of the model, using physically motivated analysis of the possible displacements to address the worst effects of the noise, applying mesh control strategies, and utilizing multiscale methods. The combination of these methods represents a new systematic approach to a class of nonrigid motion applications for which sufficiently precise and flexible FEM models can be built. The results from the skin elasticity experiments demonstrate the success of the proposed method. The model allows us to objectively detect the differences in elasticity between normal and abnormal skin. Our work demonstrates the possibility of accurate computation of point correspondences and force recovery from range image sequences containing nonrigid objects and large motion.

Image segmentation involves calculating the position of object boundaries. For scene analysis, the intent i s to di erentiate objects from clutter by means of preprocessing. The objective of this paper is to examine and discuss two morphological techniques for preprocessing and segmenting range images. A Morphological Watershed Algorithm has been studied in detail for segmenting range images. This algorithm uses a unique approach for de ning the boundaries of objects from a morphological gradient. Several sets of range images are used as input to the algorithm to demonstrate the exibility of the watershed technique and the experimental results support this approach as an e ective method for segmenting range images. Morphological image operators present another means for segmenting range images. In particular, the results from implementing gray scale morphological techniques indicate that these operators are useful for segmentation. This is made possible by converting a range image of a scene to a gray scale image representation. The result represents the umbra of the surface of the objects within the scene. By applying morphological operations to the gray v alues of the image, the operations are applied to the umbra. Each pixel represents a point of the object's umbra, thereby yielding scene segmentation. The techniques that are discussed are found to be useful for preprocessing and segmenting range images which are direct extensions to object recognition, scene analysis, and image understanding.

It is a h clear that there are m a l dBerent approaches to the problem of how to integrate visual cuea. Let us list some of the obvious poasibilities: 1 There is no active integration of visual processes. Their individual outputs ate "integrated" at the stage at which they are used, for example by a navigation system. This is the approach advocated by Brooks [1987]. While it makes sense for automatic, insectlike, visuo-motor tasks such as tracking a target or avoiding obstacles (for example, the fly' s visuo-motor system Wchardt & Poggio 1976]), it seems quite unlikely for visual perception in the wider sense. 2 The visual modules ate so tightly coupled that it is impossible to consider visual modules as separate, even to a first order approximation. This view is unattractive on epistemological, engineering, and psychophysid grounds. 3 The visual modules are coupled to each other and to the image data in a parallel fashion--each pro-represented as an array coupled to the arrays associated with the other processes. This point of view is in the tradition of Marr's 2 i-D sketch, and especially of the "intrinsic images" of Barrow and Tenenbaum (19781. Our present scheme is of this type, and exploits the machinery of Markov Random Field (MRF) models. 4 Integration of different vision modalities is taking place in a taskdependent way at specific locations--not over the whole image-and when it is needed-therefore not at all times. This approach is suggested by wychoplu.sical data on visual attentiin and by the idea of visual routines [UIlman 19841 (see also , Mahoney 119871, and ~uithoff and Mallot 119881). We are presently exploring the third of these approaches. We believe that the last two approaches are compatible with each other. In particular, visual routines may operate on maps of discontinuities such as thwe delivered by the present Vision Machine, and therefore be h a t e d after a paraUel, automatic integration stage. In real life, of course, it may be more a matt e of coexistence. We believe, in fact, that a control structure baaed on specific knowledge about the properties of the various modules, the s p d c scene and the specific task will be needed in a later version of the V i o n Machine to overview and control the MRF integration stage itself and its parameters. It is poasible that the integration stage should he much more goal-directed than what o w present methods (MRF baaed) allow. The main goal of our work is to find out whether this is true. The V i n Machine project has a number of other goals. It provides a focus for developing parallel vision algorithms and for studying how to Chapter 42 The MIT Vision Machine

The problem of place recognition appears in different mobile robot navigation problems including localization, SLAM, or change detection in dynamic environments. Whereas this problem has been studied intensively in the context of robot vision, relatively few approaches are available for threedimensional range data. In this paper, we present a novel and robust method for place recognition based on range images. Our algorithm matches a given 3D scan against a database using point features and scores potential transformations by comparing significant points in the scans. A further advantage of our approach is that the features allow for a computation of the relative transformations between scans which is relevant for registration processes. Our approach has been implemented and tested on different 3D data sets obtained outdoors. In several experiments we demonstrate the advantages of our approach also in comparison to existing techniques.

Computer vision has developed algorithms for several early vision processes, such as edge detection, stereopsis, motion, texture, and color, which give separate cues as to the distance from the viewer of three-dimensional sur-'Ibmam Po@o k The S t. & important locations in a s m e. Third, we have argued that discontinuities are ideal for integrating idormation from different visual cues.

Johnson and Hebert's spin-images have been applied to the registration of range images and object recognition with much success because they are rotation, scale, and pose invariant. In this paper we address two issues concerning spin-images, namely: (1) comparing uncompressed spinimages across large datasets is costly, and (2) a method to select the appropriate bin size and image width for spinimages is not clearly defined. Our solution to these issues is a multi-resolution method that generates a pyramid of spin-images by successively decreasing the spin-image size by powers of two. To efficiently correlate surface points, we compare spin-images in a low-to-high resolution manner. Once multi-resolution spin-images are generated for a given object, we have found that the different resolutions can also be used to compare objects that have differing or non-uniform point densities. To select the appropriate bin sizes for comparing such objects, we use the ratio of the average edge lengths of the objects. We also show preliminary results of using the pyramid to converge on the appropriate image width by traversing the pyramid in a low-to-high resolution manner looking for the highest resolution at which the fewest number of highly correlated points are found to match a given feature point.

This paper proposes a technique for object recognition using superquadric built models. Superquadrics, which are three-dimensional models suitable for part-level representation of objects, are reconstructed from range images using the recover-and-select paradigm. Using interpretation trees, the presence of an object from the model database can be hypothesized. These hypotheses are verified by projecting and re-fitting the object model to the range image of the scene which at the same time enables a better localization of the object in the scene.

Sheet-of-light range measurement is a commonly used method in todays industrial applications. This paper describes a new way of using an existing image sensor and turns it into a fast programmable sheet-of-light range sen- sor. The system is capable of measuring up to 4 Mranglels.

In many applications it is necessary to describe some experimental data with one or more geometric models. A naive approach to find multiple models consists on the sequential application of a robust regression estimator, such as RANSAC [2], and removing inliers each time that a model instance was detected. The quality of the final result in the sequential approach depends strongly on the order on which the models were. The MuSAC method proposed in this paper discovers several models at the same time, based on the consensus of each model. To reduce bad correspondences between data points and geometric models, this paper also introduces a novel distance for laser range sensors. We use the MuSAC algorithm to find models from 2D range images on cluttered environments with promising results.

This paper describes and evaluates an efficient technique that allows the fast generation of 3D triangular meshes from range images avoiding optimization procedures. Such a tool is advantageous in order to integrate range imagery into world models based on scattered representations. Furthermore, this technique can also be used as a fast preprocessing stage of registration, segmentation or recognition algorithms, owing to its abstraction capabilities that tend to eliminate redundant information. The proposed method has two stages. First, the vertices of the mesh are computed through adaptive randomized sampling of the range image based on curvature estimations. Then, the mesh is generated by triangulating the sampled vertices through an efficient 2 1 / 2 D Delaunay algorithm. The sampling process concentrates points in areas of large curvature and tends to preserve surface and orientation discontinuities. Multiresolution representations are supported in a natural way. Furthermore, this method is an inherently parallel process well suited to be implemented on high performance architectures. The proposed technique is finally evaluated with several real range images that include both free-form and polyhedral surfaces.

Range images are vision and robotics ment of fast and acquisition of 3D the other hand, in order tions may be succes!;fully constraints, such as processing. In this 1 focused on the efficient mations from range simplifying further segmentation [4]. This work focuses positions where a range to determine the surfaces with the aim of obtaining as the one described in view problem. As in el Garciat Susana VelBzquez* Science Engineering gaining popularity in the computer dommunities owing to the developlowcost range sensors that allow the information directly from a scene. On that this kind of dense representaused in fields with real time robotics, it is necessary to simplify its ne, previous work by the authors extraction of triangular approxiimages [2][5] with the aim of processing algorithms, such as image on the problem of determining the sensor should be placed in order of the objects present in a scene, a 3D model of that scene, such [3]. This is known as the next-bestseveral previous proposals, it is I Abstract occlusions.

A new technique is presented for determining the positions where a range sensor should be located to acquire the surfaces of a complex scene. The algorithm consists of two stages. The first stage applies a voting scheme that considers occlusion edges. Most of the surfaces of the scene are recovered through views computed in that way. Then, the second stage fills up remaining holes through a scheme based on visibility analysis. By leaving the more expensive visibility computations at the end of the exploration process, efficiency is increased.

This paper presents a technique for lossy compression of dense range images. Two separate compression schemes are applied. The first scheme (geometric compression) reduces redundant geometric information by generating an adaptive 3D triangular mesh that approximates the shapes present in the original range image. Geometric compression is used for obtaining an efficient representation of the range image that allows further processing. The second compression scheme (topological compression) encodes the connectivity information contained in the triangular mesh. Topological compression is used for generating a compact representation suitable to be stored or transmitted. Both compression schemes avoid costly iterative optimization algorithms. Results with real range images are presented.

This paper presents a new method for the incremental integration of overlapped range images. It is assumed that frame transformations between all pairs of views can be reliably computed. This method proceeds by progressively merging each new sensed range image with the current reconstructed model. The proposed method consists of three stages. In the first stage the overlapped regions are identified. Then, from the overlapped regions, the points that define the integration boundaries are projected over a reference plane and are triangulated over that 2D space by means of a constrained Delaunay algorithm. Finally the obtained triangulation is backprojected to the 3D range image space. These new triangular meshes represent the sewing between the meshes to be integrated. In this way a new single triangular mesh which will be used to merge with futures range images is generated. Experimental results are presented.

A new algorithm for the integration of partially overlapping range images into a triangular mesh is presented. The algorithm consists of three main steps: it locates the intersections between the range surfaces and a reference grid chosen by the user, then merges all nearly coincident and redundant intersections according to a proximity criterion, and, finally, reconstructs the merged surface(s) from the filtered intersection set. Compared with previous methods, which adopt a volumetric approach, our algorithm shows lower computational costs and improves the accuracy of the surfaces produced. It takes into account the quality of the input measurements and is able to patch small holes corresponding to the parts of the 3D scanned object that were not observed by the acquisition device. The algorithm has been tested on several datasets of range maps; graphical and numeric results are reported.

Multi-view range image integration aims at producing a single reasonable 3D point cloud. The point cloud is likely to be inconsistent with the measurements topologically and geometrically due to registration errors and scanning noise. This paper proposes a novel integration method cast in the framework of Markov random fields (MRF). We define a probabilistic description of a MRF model designed to represent not only the interpoint Euclidean distances but also the surface topology and neighbourhood consistency intrinsically embedded in a predefined neighbourhood. Subject to this model, points are clustered in aN iterative manner, which compensates the errors caused by poor registration and scanning noise. The integration is thus robust and experiments show the superiority of our MRF-based approach over existing methods.

This paper presents a 3D face photography system based on a small set of training facial range images. The training set is composed by 2D texture and 3D range images (i.e. geometry) of a single subject with different facial expressions. The basic idea behind the method is to create texture and geometry spaces based on the training set and transformations to go from one space to the other. The main goal of proposed approach is to obtain a geometry representation of a given face provided as a texture image, which undergoes a series of transformations through the texture and geometry spaces. Facial feature points are obtained by an active shape model (ASM) extracted from the 2D gray-level images. PCA then is used to represent the face dataset, thus defining an orthonormal basis of texture and range data. An input face is given by a gray-level face image to which the ASM is matched. The extracted ASM is fed to the PCA basis representation and a 3D version of the 2D input image is built. The experimental results on static images and video sequences using seven samples as training dataset show rapid reconstructed 3D faces which maintain spatial coherence similar to the human perception, thus corroborating the efficiency of our approach.

Comparison of different techniques for short-range UWB radar imaging has been performed based on experimental data. To acquire experimental data, a video-impulse radar with a specially developed antenna system with a single transmitter antenna and an array of receiver antennas has been used. Topology of the receiver array was made in such a way that the footprint of the receiver array does not have any sidelobes.

This survey focuses on recognition performed by matching models of the three-dimensional shape of the face, either alone or in combination with matching corresponding two-dimensional intensity images. Research trends to date are summarized, and challenges confronting the development of more accurate three-dimensional face recognition are identified. These challenges include the need for better sensors, improved recognition algorithms, and more rigorous experimental methodology.

This paper presents the design of a low cost largescale 3D scanner. A pan-tilt unit, a camera and a laser distance meter are combined to capture the required range and intensity data. Innovative methods for calibrating the device and fusion of the intensity and range data are introduced. Once a reliable data achieved, a parametric segmentation technique has been adopted to segment the range images into planar surfaces. Experimental results show that even for data corrupted by a large number of outliers (due to people movement in front of the imaged buildings), range and intensity data are highly consistence and segmentation algorithm has been able to properly segment coplanar areas.

In this paper, we address the problem of recovering the true underlying model of a surface while performing the segmentation. First, and in order to solve the model selection problem, we introduce a novel criterion, which is based on minimising strain energy of fitted surfaces. We then evaluate its performance and compare it with many other existing model selection techniques. Using this criterion, we then present a robust range data segmentation algorithm capable of segmenting complex objects with planar and curved surfaces. The presented algorithm simultaneously identifies the type (order and geometric shape) of each surface and separates all the points that are part of that surface. This paper includes the segmentation results of a large collection of range images obtained from objects with planar and curved surfaces. The resulting segmentation algorithm successfully segments various possible types of curved objects. More importantly, the new technique is capable of detecting the association between separated parts of a surface, which has the same Cartesian equation while segmenting a scene. This aspect is very useful in some industrial applications of range data analysis.

An objective analysis of image quality parameters was performed for six digital mammography systems. The presampled modulation transfer function (MTF), normalized noise power spectrum (NNPS), and detective quantum efficiency (DQE) for the systems were determined at different doses, for with a or target/filter combination and of additional aluminium filtration. The flat‐panel units have higher MTF and DQE in the mid to high frequency range than standard CR systems. The highest DQE, over the whole dose range, is for the slit‐scanning direct photon counting system. Dual‐side read CR can overcome the inherent x‐ray absorption and signal collection limitations of standard CR mammography, improving the low‐frequency DQE by 40%, to the same level as full‐field systems, but it does not improve the poor spatial resolution of phosphor.

Surface reconstruction from noisy point samples must take into consideration the stochastic nature of the sample. In other words, geometric algorithms reconstructing the surface or curve should not insist in following in a literal way each sampled point. Instead, they must interpret the sample as a "point cloud" and try to build the surface as passing through the best possible (in the statistical sense) geometric locus that represents the sample. This work presents two new methods to find a Piecewise Linear approximation from a Nyquist-compliant stochastic sampling of a quasi-planar C 1 curve C(u) : R → R 3 , whose velocity vector never vanishes. One of the methods articulates in an entirely new way Principal Component Analysis (statistical) and Voronoi-Delaunay (deterministic) approaches. It uses these two methods to calculate the best possible tape-shaped polygon covering the planarised point set, and then approximates the manifold by the medial axis of such a polygon. The other method applies Principal Component Analysis to find a direct Piecewise Linear approximation of C(u). A complexity comparison of these two methods is presented along with a qualitative comparison with previously developed ones. It turns out that the method solely based on Principal Component Analysis is simpler and more robust for non self-intersecting curves. For self-intersecting curves the Voronoi-Delaunay based Medial Axis approach is more robust, at the price of higher computational complexity. An application is presented in Integration of meshes originated in range images of an art piece. Such an application reaches the point of complete reconstruction of a unified mesh.

Recently, 3D models are used in many elds such as education, medical services, entertainment, art, digital archive, etc., because of the progress of computational time and demand for creating photorealistic virtual model is increasing for higher reality. In computer vision eld, a number of techniques have b e e n d e v eloped for creating the virtual model by observing the real object in computer vision eld. In this paper, we propose the method for creating photorealistic virtual model by using laser range sensor and polarization based image capture system. We capture the range and color images of the object which is rotated on the rotary table. In geometry aspect, an object surface shape is reconstructed by merging multiple range images of the object. In optical aspect, color images are captured under xed point l i g h t source. By using the reconstructed object shape and sequence of color images of the object, parameter of a re ection model are estimated in a robust manner. As a result, then, we can make photorealistic 3D model in consideration of surface re ection. The key point of the proposed method is that, rst, the di use and specular re ection components are separated from the color image sequence, and then, re ectance parameters of each re ection component are estimated separately. In separation of re ection components, we use polarization lter. This approach enables estimation of re ectance properties of real objects whose surfaces show specularity a s w ell as di usely re ected lights. The recovered object shape and re ectance properties are then used for synthesizing object images with realistic shading e ects under arbitrary illumination conditions.

Three-dimensional (3D) image acquisition systems are rapidly becoming more affordable, especially systems based on commodity electronic cameras. At the same time, personal computers with graphics hardware capable of displaying complex 3D models are also becoming inexpensive enough to be available to a large population. As a result, there is potentially an opportunity to consider new virtual reality applications as diverse as cultural heritage and retail sales that will allow people to view realistic 3D objects on home computers. Although there are many physical techniques for acquiring 3D data-including laser scanners, structured light and time-of-flight-there is a basic pipeline of operations for taking the acquired data and producing a usable numerical model. We look at the fundamental problems of range image registration, line-of-sight errors, mesh integration, surface detail and color, and texture mapping. In the area of registration we consider both the problems of finding an initial global alignment using manual and automatic means, and refining this alignment with variations of the Iterative Closest Point methods. To account for scanner line-of-sight errors we compare several averaging approaches. In the area of mesh integration, that is finding a single mesh joining the data from all scans, we compare various methods for computing interpolating and approximating surfaces. We then look at various ways in which surface properties such as color (more properly, spectral reflectance) can be extracted from acquired imagery. Finally, we examine techniques for producing a final model representation that can be efficiently rendered using graphics hardware.