Search

This paper describes our work on practical stereo vision for mobile robots using commodity hardware. The approach described in this paper is based on line segments, since those provide a lot of information about the environment, provide more depth information than point features, and are robust to image noise and colour variations. However, stereo matching with line segments is a difficult problem due to poorly localized end points and perspective distortion. Our algorithm uses integral images and Haar features for line segment extraction. Dynamic programming is used in the line segment matching phase. The resulting line segments track accurately from one frame to the next, even in the presence of noise. [ABSTRACT FROM AUTHOR]

This paper presents the fusion of monogenic signal processing and differential geometry to enable monogenic analyzing of local intrinsic 2D features of low level image data. New rotational invariant features such as structure and geometry (angle of intersection) of two superimposed intrinsic 1D signals will be extracted without the need of any steerable filters. These features are important for all kinds of low level image matching tasks in robot vision because they are invariant against local and global illumination changes and result from one unique framework within the monogenic scale-space. [ABSTRACT FROM AUTHOR]

We present a direct method to calculate a 6DoF pose change of a monocular camera for mobile navigation. The calculated pose is estimated up to a constant unknown scale parameter that is kept constant over the entire reconstruction process. This method allows a direct calculation of the metric position and rotation without any necessity to fuse the information in a probabilistic approach over longer frame sequence as it is the case in most currently used VSLAM approaches. The algorithm provides two novel aspects to the field of monocular navigation. It allows a direct pose estimation without any a-priori knowledge about the world directly from any two images and it provides a quality measure for the estimated motion parameters that allows to fuse the resulting information in Kalman Filters. We present the mathematical formulation of the approach together with experimental validation on real scene images. [ABSTRACT FROM AUTHOR]

Accurate robot detection and localisation is fundamental in applications which involve robot navigation. Typical methods for robot detection require a model of a robot. However in most applications the availability of such model can not be warranted. This paper discusses a different approach. A method is presented to localise the robot in a complex and dynamic scene based only on the information that the robot is following a previously specified movement pattern. The advantage of this method lies in the ability to detect differently shaped and differently looking robots as long as they perform the previously defined movement. The method has been successfully tested in an indoor environment. [ABSTRACT FROM AUTHOR]

In this paper, we develop a novel algorithm for navigating a mobile robot using the visual potential. The visual potential is computed from an image sequence and optical flow computed from successive images captured by the camera mounted on the robot. We assume that the direction to the destination is provided at the initial position of the robot. Using the direction to the destination, the robot dynamically selects a local pathway to the destination without collision with obstacles. The proposed algorithm does no require any knowledge or environmental maps of the robot workspace. Furthermore, this algorithm uses only a monocular uncalibrated camera for detecting a feasible region of navigation, since we apply the dominant plane detection to detect the feasible region. We present the experimental results of navigation in synthetic and real environments. Additionally, we present the robustness evaluation of optical flow computation against lighting effects and various kinds of textures. [ABSTRACT FROM AUTHOR]

With 3D vision measuring, camera calibration is necessary for extracting metric information from 2D images. We present a new algorithm for camera calibration that only requires a spatial triangle with known size. In our method, the camera is fixed while the triangle is rotated freely at one of its vertices. By taking a few (at least two) pictures of the rotating object at an identical camera view position and direction, the camera intrinsic matrix can be obtained. It advances other traditional methods in that the extrinsic parameters are separated from the intrinsic ones, which simplifies the problem with more efficiency and accuracy. Experiments also show that our method is robust and results in high resolution. [ABSTRACT FROM AUTHOR]

Vision algorithms for robot control are usually context dependent. A state based vision controller can provide both the computational and temporal context for the algorithm. A hierarchical layering enables one or more sub-objectives to be selected. To mimic human behaviour, it is argued that using fuzzy logic is better able to manage the subjective data obtained from images. Fuzzy reasoning is used to control both the transitions between states, and also to directly control the behaviour of the robot. These principles are illustrated with an autonomous guide robot. Preliminary results indicate that this approach enables complex control and vision systems to be readily constructed in a manner that is both modular and extensible. [ABSTRACT FROM AUTHOR]

We report current work on methods for robust fitting of ellipsoids to the shape of the human head in three-dimensional models built from laser scanner acquisitions. A starting ellipsoid is obtained from Principal Component Analysis from mesh vertices; those regions far from the surface of the ellipsoid are penalized (outlier rejection and/or damping). A first method consists in re-calculating iteratively the ellipsoid by PCA, adjusting scales. Another method projects the outliers into the ellipsoid. A third method uses a morphological-error approach, via the difference of the distance fields of the head and its fitting ellipsoid. The ellipsoid orientation, position and scales are then changed in search for a set of parameters minimizing the morphological error. To speedup calculations, only a few orthogonal planes are sampled and the distance field is obtained directly from the triangular mesh elements. Among the intended applications, the restoration of the skullcap is presented. [ABSTRACT FROM AUTHOR]

This paper presents vision based self-localization of tiny autonomous mobile robots in a known but highly dynamic environment. The problem covers tracking the robot position with an initial estimate to global self-localization. The algorithm enables the robot to find its initial position and to verify its location during every movement. The global position of the robot is estimated using trilateration based techniques whenever distinct landmark features are extracted. Distance measurements are used as they require fewer landmarks compared to methods using angle measurements. However, the minimum required features for global position estimation are not available throughout the entire state space. Therefore, the robot position is tracked once a global position estimate is available. Extended Kalman filter is used to fuse information from multiple heterogeneous sensors. Simulation results show that the new method that combines the global position estimation with tracking results in significant performance gain. [ABSTRACT FROM AUTHOR]

In generic object recognition, the performance of local descriptors varies from class category to another. A descriptor may have a good performance on one category and low performance on another. Combining more than one descriptor in recognition can give a solution to this problem. The choice of descriptor's type and number of descriptors to be used is then important. In this paper, we use two different types of descriptors, the Gradient Location-Orientation Histogram (GLOH) and simple color descriptor, for generic object recognition. Boosting is used as the underlying learning technique. The recognition model achieves a performance that is comparable to or better than that of state-of-the-art approaches. [ABSTRACT FROM AUTHOR]

We introduce the problem of autonomous trail following without waypoints and present a vision- and ladar-based system which keeps to continuous hiking and mountain biking trails of relatively low human difficulty. Using a RANSAC-based analysis of ladar scans, trail-bordering terrain is classified as belonging to one of several major types: flat terrain, which exhibits low height contrast between on- and off-trail regions; thickly-vegetated terrain, which has corridor-like structure; and forested terrain, which has sparse obstacles and generally lower visual contrast. An adaptive color segmentation method for flat trail terrain and a height-based corridor-following method for thick terrain are detailed. Results are given for a number of autonomous runs as well as analysis of logged data, and ongoing work on forested terrain is discussed. [ABSTRACT FROM AUTHOR]

Person tracking is a key requirement for modern service robots. But methods for robot vision have to fulfill several constraints: they have to be robust to errors evoked by noisy sensor data, they have to be able to work under real-world conditions, and they have to be fast and computationally inexpensive. In this paper we present an approach for tracking the position of a person in 3D based on a particle filter. In our framework, each particle represents a hypothesis for the 3D position, velocity and size of the person's head being tracked. Two cameras are used for the evaluation of the particles. The particles are weighted by projecting them onto the camera image and applying a color-based perception model. This model uses skin color cues for face tracking and color histograms for body tracking. In contrast to feature-based approaches, our system even works when the person is temporary or partially occluded. [ABSTRACT FROM AUTHOR]

Quality of 3D modelling in means of accuracy and level of detail is based fundamentally on the parameters of the imaging system. All aspects from camera design over calibration up to processing have to be regarded and to be coordinated. The whole system chain has to be realized and controlled. The German Aerospace Center developed a new generation of line cameras called MFC (multi functional camera), which can be applied for spaceborne, airborne and terrestrial data acquisition. Its core is a CCD line module consisting of the imaging detector and the front end electronics. Any desired number of such modules can be combined to build up a custom specific imaging system. Flexible usage of different CCD sensors, USB2 interface, real time noise reduction and image compression are the main parameters of this camera. The application of an inertial measurement unit allows data acquisition with moving platforms. This paper gives an overview about the system parameters. All steps of data processing are introduced and examples of 3D models will be given. State of the art stereo matching algorithms are of prime importance, therefore the applied semi-global matching is described shortly. The quality of data products will be evaluated. ... [ABSTRACT FROM AUTHOR]

In recent years a lot of versatile robots have been developed to work in environments with human. However they are not sufficiently flexible nor safe in terms of interaction with human. In our approach we focused on hand and eye coordination in order to establish a flexible robot control, in which a robot recognizes its environment from the input camera images and equips a soft contacting strategy by impedance control. To recognize the environment, we adopt a method to reconstruct motion from a sequence of monocular images by using a pair of parallel straight line segments, which enables us to obtain linear equations to solve the problem. On the other hand the impedance control strategy conveys a flexible interaction between robots and humans. The strategy can be considered as a passive force control, when something contacts the end-effector of the robot. In order to avoid a collision, we introduce a virtual impedance control which can generate force prior to the contact. Neural networks (hereafter: NN) learning is used to decide parameters for impedance control, in which NNs can obtain parameters during the motion (aka: online learning). The validity of the proposed method was verified through experiments with a multijoint robot manipulator. [ABSTRACT FROM AUTHOR]

The history of stereo analysis of pairs of images dates back more than one hundred years, but stereo analysis of image stereo sequences is a fairly recent subject. Sequences allow time-propagation of results, but also come with particular characteristics such as being of lower resolution, or with less contrast. This article discusses the application of belief propagation (BP), which is widely used for solving various low-level vision problems, for the stereo analysis of night-vision stereo sequences. For this application it appears that BP often fails on the original frames for objects with blurry borders (trees, clouds...). In this paper, we show that BP leads to more accurate stereo correspondence results if it is applied on edge images, where we have decided for the Sobel edge operator, due to its time efficiency. We present the applied algorithm and illustrate results (without, or with prior edge processing) on seven, geometrically rectified night-vision stereo sequences (provided by Daimler research, Germany). [ABSTRACT FROM AUTHOR]

We will demonstrate the sensor and data fusion concept of the 2007 DARPA Urban Challenge vehicle assembled by Team CarOLO, Technische Universität Braunschweig. The perception system is based on a hybrid fusion concept, combining object and grid based approaches in order to comply with the requirements of an urban environment. A variety of sensor systems and technologies is applied, providing a 360 degree view area around the vehicle. Within the object based subsystem, obstacles (static and dynamic) are tracked using an Extended Kalman Filter capable of tracking arbitrary contour shapes. Additionally, the grid based subsystem extracts drivability information about the vehicle's driveway by combining the readings of laser scanners, a mono and a stereo camera system using a Dempster-Shafer based data fusion approach. ... [ABSTRACT FROM AUTHOR]

This paper presents a vision-based color segmentation algorithm suitable for urban environments that separates an image into areas of drivable and non-drivable terrain. Assuming that a part of the image is known to be drivable terrain, other parts of the image are classified by comparing the Euclidean distance of each pixel's color to the mean colors of the drivable area in real-time. Moving the search area depending on each frame's result ensures temporal consistency and coherence. Furthermore, the algorithm classifies artifacts such as white and yellow lane markings and hard shadows as areas of unknown drivability. The algorithm was thoroughly tested on the autonomous vehicle 'Caroline', which was a finalist in the 2007 DARPA Urban Challenge. [ABSTRACT FROM AUTHOR]

This paper reports on AnnieWAY, an autonomous vehicle that is capable of driving through urban scenarios and that has successfully entered the finals of the DARPA Urban Challenge 2007 competition. After describing the main challenges imposed and the major hardware components, we outline the underlying software structure and focus on selected algorithms. A recent laser scanner plays the prominent role in the perception of the environment. It measures range and reflectivity for each pixel. While the former is used to provide 3D scene geometry, the latter allows robust lane marker detection. Mission and maneuver selection is conducted via a concurrent hierarchical state machine that specifically ascertains behavior in accordance with California traffic rules. We conclude with a report of the results achieved during the competition. [ABSTRACT FROM AUTHOR]

This paper discusses a trajectory based recognition algorithm to expand the common approaches for atypical event detection in multi-object traffic scenes and to obtain area-based types of information (e.g. maps of speed patterns, trajectory curvatures or erratic movements). Different views of the same area by more than one camera sensor are necessary, because of the typical limitations of single camera systems, resulting from occlusions by other cars, trees and traffic signs. Furthermore, distributed cooperative multi-camera system (MCS) enables a significant enlargement of the observation area. The fusion of object data from different cameras is done by a multi-target tracking approach. This approach opens up opportunities to identify and specify traffic objects, their location, speed and other characteristic object information. New and consolidated information of traffic participants is derived from the system. [ABSTRACT FROM AUTHOR]

Automatic road-signs recognition is becoming a part of Driver Assisting Systems which role is to increase safety and driving comfort. In this paper we present architecture and obtained results of the Polish project on the road-signs recognition system. A holistic approach was undertaken, focusing on interaction of different components of the system. [ABSTRACT FROM AUTHOR]

Blind spots usually lead to difficulties for drivers to maneuver their vehicles in complicated environments, such as garages, parking spaces, or narrow alleys. This paper presents a vision system which can assist drivers by providing the panoramic image of vehicle surroundings in a bird's-eye view. In the proposed system, there are six fisheye cameras mounted around a vehicle so that their views cover the whole surrounding area. Parameters of these fisheye cameras were calibrated beforehand so that the captured images can be dewarped into perspective views for integration. Instead of error-prone stereo matching, overlapping regions of adjacent views are stitched together by aligning along a seam with dynamic programming method followed by propagating the deformation field of alignment with Wendland functions. In this way the six fisheye images can be integrated into a single, panoramic, and seamless one from a look-down viewpoint. Our experiments clearly demonstrate the effectiveness of the proposed image-stitching method for providing the bird's eye view vision for vehicle surrounding monitoring. [ABSTRACT FROM AUTHOR]

The watchman route problem (WRP) was first introduced in 1988 and is defined as follows: How to calculate a shortest route completely contained inside a simple polygon such that any point inside this polygon is visible from at least one point on the route? So far the best known result for the WRP is an ${\cal O}(n^3 \log n)$ runtime algorithm (with inherent numerical problems of its implementation). This paper gives an $\kappa(\varepsilon)\times {\cal O}(kn)$ approximate algorithm for WRP by using a rubberband algorithm, where n is the number of vertices of the simple polygon, k the number of essential cuts, ε the chosen accuracy constant for the minimization of the calculated route, and κ(ε) equals the length of the initial route minus the length of the calculated route, divided by ε. [ABSTRACT FROM AUTHOR]

Kruppa equation based camera self-calibration is one of the classical problems in computer vision. Most state-of-the-art approaches directly solve the quadratic constraints derived from Kruppa equations, which are computationally intensive and difficult to obtain initial values. In this paper, we propose a new initialization algorithm by estimating the unknown scalar in the equation, thus the camera parameters can be computed linearly in a closed form and then refined iteratively via global optimization techniques. We prove that the scalar can be uniquely recovered from the infinite homography and propose a practical method to estimate the homography from a physical or virtual plane located at a far distance to the camera. Extensive experiments on synthetic and real images validate the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

We propose an accurate method for image matching in scenes including repetitive patterns like buildings, walls, and so on. We construct our matching method with two phases: matching between the elements of repetitive regions; matching between the points in the remained regions. We first detect the elements of repetitive patterns in each image and find matches between the elements in the regions without using any descriptors depended on a view-point. We then find matches between the points in the remained regions of the two images using the informations of the detected matches. The advantage of our method is to use an efficient matching information in the repetitive patterns. We show the effectiveness of our method by real image examples. [ABSTRACT FROM AUTHOR]

We present a highly efficient method for estimating the structure and motion from image sequences taken by uncalibrated cameras. The basic principle is to do projective reconstruction first followed by Euclidean upgrading. However, the projective reconstruction step dominates the total computational time, because we need to solve eigenproblems of matrices whose size depends on the number of frames or feature points. In this paper, we present a new algorithm that yields the same solution using only matrices of constant size irrespective of the number of frames or points. We demonstrate the superior performance of our algorithm, using synthetic and real video images. [ABSTRACT FROM AUTHOR]

To be effective in the human world robots must respond to human emotional states. This paper focuses on the recognition of the six universal human facial expressions. In the last decade there has been successful research on facial expression recognition (FER) in controlled conditions suitable for human-computer interaction [1,2,3,4,5,6,7,8]. However the human-robot scenario presents additional challenges including a lack of control over lighting conditions and over the relative poses and separation of the robot and human, the inherent mobility of robots, and stricter real time computational requirements dictated by the need for robots to respond in a timely fashion. Our approach imposes lower computational requirements by specifically adapting model-based techniques to the FER scenario. It contains adaptive skin color extraction, localization of the entire face and facial components, and specifically learned objective functions for fitting a deformable face model. Experimental evaluation reports a recognition rate of 70% on the Cohn-Kanade facial expression database, and 67% in a robot scenario, which compare well to other FER systems. [ABSTRACT FROM AUTHOR]

The objective of this research is to apply markerless Augmented Reality (AR) techniques to aid in the visualisation of robotic helicopter related tasks. Conventional robotic AR applications work well with markers in prepared environments but are infeasible in outdoor settings. In this paper, we present preliminary results from a real time markerless AR system for tracking natural features in an agricultural scene. By constructing a virtual marker under a known initial configuration of the robotic helicopter, camera and the ground plane, the camera pose can be continuously tracked using the natural features from the image sequence to perform augmentation of virtual objects. The experiments are simulated on a mock-up model of an agricultural farm and the results show that the current AR system is capable of tracking the camera pose accurately for translational motions and roll rotations. Future work includes reducing jitter in the virtual marker vertices to improve camera pose estimation accuracy for pitch and yaw rotations, and implementing feature recovery algorithms. [ABSTRACT FROM AUTHOR]

Finding correspondences between sensor measurements obtained at different places is a fundamental task for an autonomous mobile robot. Most matching methods search correspondences between salient features extracted from such measurements. However, finding explicit matches between features is a challenging and expensive task. In this paper we build a local map using a stereo head aided by sonars and propose a method for aligning local maps without searching explicit correspondences between primitives. From objects found by the stereo head, an object probability density distribution is built. Then, the Gauss-Newton algorithm is used to match correspondences, so that, no explicit correspondences are needed. [ABSTRACT FROM AUTHOR]

We consider the epipolar geometry between two omnidirectional images acquired from a single viewpoint catadioptric vision sensor with parabolic mirror. This work in particular deals with the estimation of the respective epipoles. We use conformal geometric algebra to show the existence of a 3× 3 essential matrix, which describes the underlying epipolar geometry. The essential matrix is preferable to the 4× 4 fundamental matrix, which comprises the fixed intrinsic parameters, as it can be estimated from less data. We use the essential matrix to obtain a prior for a stochastic epipole computation being a key aspect of our work. The computation uses the well-tried amalgamation of a least-squares adjustment (LSA) technique, called the Gauss-Helmert method, with conformal geometric algebra. The imaging geometry enables us to assign distinct uncertainties to the image points which justifies the considerable advantage of our LSA method over standard estimation methods. Next to the stochastically optimal position of the epipoles, the method computes the rigid body motion between the two camera positions. In addition, our text demonstrates the effortlessness and elegance with which such problems integrate into the framework of geometric algebra. [ABSTRACT FROM AUTHOR]

This work deals with modeling and tracking of mechanical systems which are given as kinematic chains with restricted degrees of freedom. Such systems may involve many joints, but due to additional restrictions or mechanical properties the joints depend on each other. So-called closed-chain or parallel manipulators are examples for kinematic chains with additional constraints. Though the degrees of freedom are limited, the complexity of the dynamic equations increases rapidly when studied analytically. In this work, we suggest to avoid this kind of analytic integration of interconnection constraints and instead to model them numerically via soft constraints. [ABSTRACT FROM AUTHOR]

Efficient stereo matching algorithms search for corresponding points along the epipolar lines defined by the camera properties and configuration. In real cameras, lens distortions and misaligned cameras lead to sloping curved epipolar lines on the raw images. To minimize latencies in a system designed to produce 3D environment maps in real time, positions on the epipolar lines must be computed rapidly and efficiently. We describe the rectification components of a real-time high resolution stereo system in which two high resolution cameras are connected via FireWire links to an FPGA which removes lens distortion and misalignment errors with minimal latency. Look up tables and interpolations are used to provide a balance between high speed and accuracy. After analysing the loss in accuracy from this approach, we show that, for lenses with significant distortions, tables which contain an entry for every 16th pixel provide accurate interpolated intensities. For better (lower distortion) lenses, the lookup tables can be even smaller. The latency is several scan lines and is determined by the lens distortions and camera misalignments themselves. [ABSTRACT FROM AUTHOR]

Stereo vision is a key technology for understanding natural scenes. Most research concentrates on single image pairs. However, in robotic and intelligent vehicles applications image sequences have to be analyzed. The paper shows that an appropriate evaluation in time gives much better results than classical frame-by-frame reconstructions. We start with the state-of-the art in real-time stereo analysis and describes novel techniques to increase the sub-pixel accuracy. Secondly, we show that static scenes seen from a moving observer can be reconstructed with significantly higher precision, if the stereo correspondences are integrated over time. Finally, an optimal fusion of stereo and optical flow, called 6D-Vision, is described that directly estimates position and motion of tracked features, even if the observer is moving. This eases the detection and tracking of moving obstacles significantly. [ABSTRACT FROM AUTHOR]

Intelligent vehicle systems need to distinguish which objects are moving and which are static. A static concrete wall lying in the path of a vehicle should be treated differently than a truck moving in front of the vehicle. This paper proposes a new algorithm that addresses this problem, by providing dense dynamic depth information, while coping with real-time constraints. The algorithm models disparity and disparity rate pixel-wise for an entire image. This model is integrated over time and tracked by means of many pixel-wise Kalman filters. This provides better depth estimation results over time, and also provides speed information at each pixel without using optical flow. This simple approach leads to good experimental results for real stereo sequences, by showing an improvement over previous methods. [ABSTRACT FROM AUTHOR]

This paper introduces a new algorithm for computing multi-resolution optical flow, and compares this new hierarchical method with the traditional combination of the Lucas-Kanade method with a pyramid transform. The paper shows that the new method promises convergent optical flow computation. Aiming at accurate and stable computation of optical flow, the new method propagates results of computations from low resolution images to those of higher resolution. The resolution of images increases this way for the sequence of images used in those calculations. The given input sequence of images defines the maximum of possible resolution. [ABSTRACT FROM AUTHOR]

In this paper, we present a specific use of the Particle-based Belief Propagation (PBP) algorithm as an approximation scheme for the joint distribution over many random variables with very large or continuous domains. After formulating the problem to be solved as a probabilistic graphical model, we show that by running loopy Belief Propagation on the whole graph, in combination with an MCMC method such as Metropolis-Hastings sampling at each node, we can approximately estimate the posterior distribution of each random variable over the state space. We describe in details the application of PBP algorithm to the problem of sparse Structure from Motion and the dense Stereo Vision with unknown camera constraints. Experimental results from both cases are demonstrated. An experiment with a synthetic structure from motion arrangement shows that its accuracy is comparable with the state-of-the-art while allowing estimates of state uncertainty in the form of an approximate posterior density function. [ABSTRACT FROM AUTHOR]

The tumor segmentation in Breast MRI image is difficult due to the complicated galactophore structure. The work in this paper attempts to accurately segment the abnormal breast mass in MRI(Magnetic resonance imaging) Images. The ROI (Region of Interest) is segmented using a novel DP (Dynamic Programming) based optimal edge detection technique. DP is an optimal approach in multistage decision-making. The method presented in this paper processes the object image to get the minimum cumulative cost matrix combining with LUM nonlinear enhancement filter, Gaussian preprocessor, non-maximum suppression and double-threshold filtering, and then trace the whole optimal edge. The experimental results show that this method is robust and efficient on image edge detection and can segment the breast tumor area more accurately. [ABSTRACT FROM AUTHOR]

Based on three dimensional echocardiography sequences, a new technique using connected equi-length curve angle chain (CELCAC) model for reconstruction and dynamic analysis of mitral annulus is proposed in this paper. Firstly, the boundary points of mitral annulus of the mitral annulus is extracted by interactive method and ordered according to their positions. Then, the three dimensional mitral annulus visualization model is established based on non-uniform rational B - spline algorithm. Finally, dynamic analysis of mitral annulus presented by the CELCAC model is suggested. The presentation is invariant to rotation, scaling, and translation. Results show that the reconstruction and analysis method proposed in this paper is feasible to assess the geometry and analyze the motion of mitral annulus. [ABSTRACT FROM AUTHOR]

The problem of motion in PET/CT studies results from the fact that the PET images are formed over an elongated period of time whereas the CT acquisition takes only a few seconds. Additionally, the CT images are also used for attenuation correction of the PET data. The spatial discrepancy between the two sets of images, however, may lead to wrong attenuation and thus to wrong quantification of the radioactive uptake. We present a solution to this problem by respiratory-gating the PET data and correcting the PET images for motion with a multiresolution optical flow algorithm. Our algorithm is based on the combined local and global optical flow algorithm with modifications to allow for discontinuity preservation across organ boundaries and for application to 3D volume sets. [ABSTRACT FROM AUTHOR]

This paper proposes an novel approach for automatic gridding of cDNA microarray images based on histogram representation. The approach constructs histogram representation to characterize microarray images and use it, instead of raw signal intensities used in previous gridding methods, to identify spots. The histogram representation can efficiently reduce noise and the influence from low raw signal intensities and local contaminations. The proposed approach is successfully tested on different types of microarray images and is compared with several previous algorithms. [ABSTRACT FROM AUTHOR]

Tongue diagnosis is a distinctive and essential diagnostic measure in Traditional Chinese Medicine (TCM), and chromatic information is its most decisive characteristic which is utilized to unearth pathological changes for identifying diseases. In this paper, a computerized medical biometrics scheme is established which classify all pixels in tongue into various color classes. For train specimens, both a forward and a backward selection are employed to pick up the correct labeled pixel specimens and screen out the wrong, and then various pixels of diversified colors are classified for the tongue color analysis. The experimental outcomes are more applicable to tongue color analysis. [ABSTRACT FROM AUTHOR]

Today, tomographic images are used in medical applications more and more. To support physicians in diagnosis and treatment, a registration of two images taken at different points in time or under different conditions is needed. As the structure of the vessel or airway trees is relatively stable between two image acquisitions, they provide a good basis for the automatic determination of landmarks. In this work, a hierarchical tree search algorithm is proposed, which efficiently computes a matching between branchpoints of anatomical trees, which can be used as landmarks for an elastic registration. The algorithm is designed to be general and robust in order to be applicable to a variety of different datasets, which are acquired by different sensors or under different conditions. The validation of the algorithm against manually created ground truth data leads to a 80.9% rate of correctly matched branchpoints. Allowing a tolerance of 5 mm, the rate increases to 89.9%. The runtime for 50-700 vertices is about 1-45 seconds. [ABSTRACT FROM AUTHOR]

Tongue image matching is an important part for a tongue diagnosis system. Unlike common pattern recognition problems, it is hard to define the ground truth for tongue image matching because visual inspection of tongue by doctors is determined by the experience and knowledge of them. Here we propose to use, Mean Rank, as an objective and scientific criterion to evaluate matching performance. Instigating from color demosaicking, a new color texture operator, Primary Difference Signal Local Binary Pattern is proposed. The matching performance is evaluated on color, gray-scale and color texture, and fusion of color and texture features. [ABSTRACT FROM AUTHOR]

Now a days all medical imaging equipments give output as digital image and non-invasive techniques are becoming cheaper, the database of images is becoming larger. This archive of images increases up to significant size and in telemedicine-based applications the storage and transmission requires large memory and bandwidth respectively. There is a need for compression to save memory space and fast transmission over internet and 3G mobile with good quality decompressed image, even though compression is lossy. This paper presents a novel approach for designing enhanced vector quantizer, which uses Kohonen's Self Organizing neural network. The vector quantizer (codebook) is designed by training with a neatly designed training image and by selective training approach .Compressing; images using it gives better quality. The quality analysis of decompressed images is evaluated by using various quality measures along with conventionally used PSNR. [ABSTRACT FROM AUTHOR]

This paper presents a gradient direction edge enhancement based contour (GDEEBC) detector to segment the nucleus and cytoplasm from each cervical smear image. In order to eliminate noise from the image, this paper proposes a trim-meaning filter that can effectively remove impulse and Gaussian noise but still preserve the edge sharpness of an object. In addition, a bi-group enhancer is proposed to make a clear-cut separation for the pixels lying between two objects. Finally, a gradient direction (GD) enhancer is presented to suppress the gradients of noise and to brighten the gradients of object contours as well. The experimental results show that all the techniques proposed above have impressive performances. In addition to cervical smear images, these proposed techniques can also be utilized in object segmentation of other types of images. [ABSTRACT FROM AUTHOR]

In single photon emission computed tomography (SPECT), pinhole and cone-beam collimators are used for small object imaging. It is accepted that a pinhole collimator should be used if the object is very small, and the cone-beam collimator should be used if the object is larger. This paper presents criteria to determine which collimation geometry is more advantageous in terms of spatial resolution and detection sensitivity. [ABSTRACT FROM AUTHOR]

This paper presents a new interactive method for analyzing 4D Computed Tomography (4DCT) datasets for patient care and research. Deformable registration algorithms are commonly used to observe the trajectories of individual voxels from one respiratory phase to another. Our system provides users with the capability to visualize these trajectories while simultaneously rendering anatomical volumes, which can greatly improve the accuracy of deformable registration. [ABSTRACT FROM AUTHOR]

Dual-energy mammography can suppress the contrast between adipose and glandular tissues and improve the detectability of microcalcifications (MCs). In the published papers, MCs were calibrated by aluminum and identified by their thickness. However, the variety of compositions of MCs causes the variety of attenuation differences between MCs and MC calibration material which bring about huge calculation errors. In our study, we selected calcium carbonate and calcium phosphate as the most suitable MC calibration materials and the correction coefficient was reasonably determined. Area density was used as MC indicator instead of thickness. Therefore, the calculation errors from MC calibration materials can be reduced a lot and the determination of MCs will become possible. [ABSTRACT FROM AUTHOR]

Osteoporosis is a disease characterized by decreased bone mass and progressive deterioration of the microstructure, affecting both mineral density and bone's fragility. Current diagnoses are only measuring apparent bone mineral density (AppBMD). Using our newly developed scanning confocal acoustic diagnostic (SCAD) system, we evaluated the ability of quantitative ultrasound in noninvasively predicting bone's quantity and quality on 19 human cadaver calcanei. Results show that ultrasound attenuation image on intact calcaneus represents bone mass distribution. High correlation (R=0.82) exists between SCAD determined broadband ultrasound attenuation (BUA) and DXA determined AppBMD at the calcaneus, as well as in the AppBMD result at femoral neck (R=0.81). SCAD determined BUA and ultrasound velocity (UV) are highly correlated with the micro-CT and mechanical testing determined bone quantity and quality parameters. These results suggest that image-based quantitative ultrasound is able to identify ROI and predict both bone mass and strength. [ABSTRACT FROM AUTHOR]

Osteoporosis is a systemic skeletal disease, which is characterized by low bone mass and micro-architectural deterioration of bone tissue, leading to bone fragility. Finding an effective method for prevention and early diagnosis of the disease is very important. Several parameters, including broadband ultrasound attenuation (BUA), speed of sound (SOS), and stiffness index (STI), have been used to measure the characteristics of bone tissues. In this paper, we proposed a method, namely modified contour deformable model (MCDM), bases on the active contour model (ACM) and active shape model (ASM) for automatically detecting the calcaneus contour from quantitative ultrasound (QUS) parametric images. The results show that the difference between the contours detected by the MCDM and the true boundary for the phantom is less than one pixel. By comparing the phantom ROIs, significant relationship was found between contour mean and bone mineral density (BMD) with R=0.99. The influence of selecting different ROI diameters (12, 14, 16 and 18 mm) and different region-selecting methods, including fixed region (ROIfix), automatic circular region (ROIcir) and calcaneal contour region (ROIanat), were evaluated for testing human subjects. Measurements with large ROI diameters, especially using fixed region, result in high position errors (10-45%). The precision errors of the measured ultrasonic parameters for ROIanat are smaller than ROIfix and ROIcir. In conclusion, ROIanat provides more accurate measurement of ultrasonic parameters for the evaluation of osteoporosis and is useful for clinical application. [ABSTRACT FROM AUTHOR]

Magnetic resonance imaging and spectroscopy (MRI/MRS) plays a unique role in multiple sclerosis (MS) evaluation, because of its ability to provide both high image contrast and significant chemical change among brain tissues. The image contrast renders the possibility of quantifying the tissue volumetric and texture variations, e.g., cerebral atrophy and progressing speed, reflecting the ongoing destructive pathologic processes. Any chemical change reflects an early sign of pathological alteration, e.g., decreased N-acetyl aspartate (NAA) in lesions and normal appearing white matter, related to axonal damage or dysfunction. Both MRI and MRS encounter partial volume (PV) effect, which compromises the quantitative capability, especially for MRS. This work aims to develop a statistical framework to segment the tissue mixtures inside each image element, eliminating theoretically the PV effect, and apply the framework to the evaluation of MS with cognitive impairment. The quantitative measures from MRI/MRS neuroimaging are strongly correlated with the qualitative neuropsychological scores of Brief Repeatable Battery (BRB) test on cognitive impairment, demonstrating the usefulness of the PV image segmentation framework in this clinically significant problem. [ABSTRACT FROM AUTHOR]