Search

Casting is one of the most important processes in the manufacture of parts for various kinds of industries, among which the automotive industry stands out. Like every manufacturing process, there is the possibility of the occurrence of defects in the materials from which the parts are made, as well as of the appearance of faults during their operation. One of the most important tools for verifying the integrity of cast parts is radioscopy. This paper presents pattern recognition methodologies in radioscopic images of cast automotive parts for the detection of defects. Image processing techniques were applied to extract features to be used as input of the pattern classifiers developed by artificial neural networks. To estimate the accuracy of the classifiers, use was made of random selection techniques with sample reposition (Bootstrap technique) and without sample reposition. This work can be considered innovative in that field of research, and the results obtained motivate this paper. [ABSTRACT FROM AUTHOR]

In this article is proposed a new approach to automatic fingerprint verification that is not based on the standard ridge-minutiae-based framework, but in a general-purpose wide baseline matching methodology. Instead of detecting and matching the standard structural features, in the proposed approach local interest points are detected in the fingerprint, then local descriptors are computed in the neighborhood of these points, and afterwards these descriptors are compared using local and global matching procedures. The final verification is carried out by a Bayes classifier. It is important to remark that the local interest points do not correspond to minutiae or singular points, but to local maxima in a scale-space representation of the fingerprint images. The proposed system has 4 variants that are validated using the FVC2004 test protocol. The best variant, which uses an enhanced fingerprint image, SDoG interest points and SIFT descriptors, achieves a FRR of 20.9% and a FAR of 5.7% in the FVC2004-DB1 test database, without using any minutia or singular points' information. [ABSTRACT FROM AUTHOR]

The study of tree-rings is a common task in dendrology. Usually the rings deliver information about the age of the tree, historic climate conditions and forest densities. Many different techniques exist to perform the tree-ring detection, but they commonly are semi-automatic. The main idea of this work is to propose an automatic process for the tree-ring detection and compare it with a manual detection made by an expert in dendrology. The proposed technique is based on a variant of the Generalized Hough Transform (GHT) created using a very simple growing model of the tree. The presented automatic algorithm shows tolerance to textured and very noisy images, giving a good tree-ring recognition in most of the cases. In particular, it correctly detects the 80% of the tree-rings in our sample database. [ABSTRACT FROM AUTHOR]

This paper presents a novel spatial-invariant filtering method for rendering focus effects without aliasing artifacts from undersampled light fields. The presented method does not require any scene analysis such as depth estimation and feature matching. First, we generate a series of images focused on multiple depths by using the conventional synthetic aperture reconstruction method and treat them as a 3D image. Second we convert it to the alias-free 3D image. This paper shows this conversion can be achieved simply by a 3D filtering in the frequency domain. The proposed filter can also produce depth-of-field effects. [ABSTRACT FROM AUTHOR]

This paper presents a study about the application of Zernike invariants to content-based Image Retrieval for 2D color images. Zernike invariants have been chosen because of their good performance for object recognition. Taking into account the good results achieved in previous CBIR experiments with color based primitives using a multiresolution representation of the visual contents, this paper presents the application of a wavelet transform to the images in order to obtain a multiresolution representation of the shape based features studied. Experiments have been performed using two databases: the first one is a small self-made 2D color database formed by 298 RGB images and a test set with 1655 query images that has been used for preliminary tests; the second one is Also experiments using the Amsterdam Library of Object Images (ALOI), a free access database. Experimental results show the feasibility of this new approach. [ABSTRACT FROM AUTHOR]

Accurate object recognition based on image processing is required in embedded applications, where real-time processing is expected to incorporate accurate recognition. To achieve accurate real-time object recognition, an accurate recognition algorithm that can be quickened by parallel implementation and a processing system that can execute such algorithms in real-time are necessary. In this paper, we implemented an accurate recognition scheme in parallel that consists of boosting-based detection and histogram-based tracking on a Cell Broadband Engine (Cell), one of the latest high performance embedded processors. We show that the Cell can achieve real-time object recognition on QVGA video at 22 fps with three targets and 18 fps with eight targets . Furthermore, we constructed a real-time object recognition system using SONY® Playstation 3, one of the most widely used Cell platforms, and demonstrated face recognition with it. [ABSTRACT FROM AUTHOR]

This paper presents the architecture of an image administration system that supports the medical practice in tasks such as teaching, diagnosis and telemedicine. The proposed system has a multi-tier, web-based architecture and supports content-based retrieval. The paper discusses the design aspects of the system as well as the proposed content-based retrieval approach. The system was tested with real pathology images to evaluate its performance, reaching a precision rate of 67%. The detailed results are presented and discussed. [ABSTRACT FROM AUTHOR]

Image replica detection becomes very active research field recently as the electronic device such as the digital camera which generates digital images spreads out rapidly. As huge amount of digital images leads to severe problems like copyright protection, the necessity of replica detection system gets more and more attention. In this paper, we propose a new fast image replica detector based on concentric circle partition method. The proposed algorithm partitions image into concentric circle with fixed angle from image center position outwards. From these partitioned regions, total of four features are extracted. They are average intensity distribution and its difference, symmetrical difference distribution and circular difference distribution in bitstring type. To evaluate the performance of the proposed method, pair-wise independence test and accuracy test are applied. We compare the duplicate detection performance of the proposed algorithm with that of the MPEG-7 visual descriptors. From experimental results, we can tell that the proposed method shows very high matching speed and high accuracy on the detection of replicas which go through many modification from the original. Because we use the hash code as the image signature, the matching process needs very short computation time. And the proposed method shows 97.6 under 1 part per million false positive rate. [ABSTRACT FROM AUTHOR]

Shape is a fundamental image feature and belongs to one of the most important image features used in Content-Based Image Retrieval. This feature alone provides capability to recognize objects and retrieve similar images on the basis of their contents. In this paper, we propose a neural network-based shape retrieval system in which moment invariants and Zernike moments are used to form a feature vector. k-means clustering is used to group correlated and similar images in an image collection into k disjoint clusters whereas neural network is used as a retrieval engine to measure the overall similarity between the query and the candidate images. The neural network in our scheme serves as a classifier such that the moments are input to it and its output is one of the k clusters that has the largest similarity to the query image. [ABSTRACT FROM AUTHOR]

Content-based image retrieval (CBIR) is currently limited because of the lack of representational power of the low-level image features, which fail to properly represent the actual contents of an image, and consequently poor results are achieved with the use of this sole information. Spatial relations represent a class of high-level image features which can improve image annotation. We apply spatial relations to automatic image annotation, a task which is usually a first step towards CBIR. We follow a probabilistic approach to represent different types of spatial relations to improve the automatic annotations which are obtained based on low-level features. Different configurations and subsets of the computed spatial relations were used to perform experiments on a database of landscape images. Results show a noticeable improvement of almost 9% compared to the base results obtained using the k-Nearest Neighbor classifier. [ABSTRACT FROM AUTHOR]

This paper proposes a fuzzy reasoning interpolation method for video deinterlacing. We propose edge detection parameters to measure the amount of entropy in the spatial and temporal domains. The shape of the membership functions is designed adaptively, according to those parameters and can be utilized to determine edge direction. Our proposed fuzzy edge direction detector operates by identifying small pixel variations in nine orientations in each domain and uses rules to infer the edge direction. We employ a Bayesian network, which provides accurate weightings between the proposed deinterlacing method and common existing deinterlacing methods. It successively builds approximations of the deinterlaced sequence by weighting interpolation methods. The results of computer simulations show that the proposed method outperforms a number of methods in the literature. [ABSTRACT FROM AUTHOR]

Recently, many researches on frame skipping are conducted to reduce temporal redundancy in video frames. As a simple method, fixed frame skipping (FFS) adjusts frame rate by skipping frame at regular intervals. To overcome the poor performance of FFS, variable frame skipping (VFS) has been introduced to exploit the temporal dependency between frames. In this paper, scene-adaptive key frame selection method with low complexity is proposed. The proposed method performed about 20 percent better in complexity with the better visual quality than the conventional video encoding. As a preprocessing method, the proposed technology can be used with any conventional video codec. [ABSTRACT FROM AUTHOR]

Distributed Video Coding (DVC) is a promising coding solution for some emerging applications, where the encoder complexity, power consumption or memory requirements are constraint the system resources. Current approaches to DVC focus on improving the performance of the Wyner-Ziv coding by improving the quality of the reconstructed side information or by improving the quality of channel codes. Up to date, no attention has been paid to the problem of key frames coding where a low-encoding complexity scenario is also needed. This work focuses on key frames coding in its effect to the Wyner-Ziv frames decoding aiming to implement a very low-complexity Turbo Trellis Coded Modulation (TTCM) based DVC architecture. In this paper, we propose a new key frame coding scheme which has very low complexity and memory requirements for the TTCM based distributed video codec. Results show that the proposed intra frame codec for key frame coding outperforms the JPEG2000 and the Intra H.264 AVC codecs in terms of encoding-time and memory requirements, with better RD performance. [ABSTRACT FROM AUTHOR]

This paper deals with a transformation of raster grey-scale images into a geometric representation by the Delaunay triangulation. It discusses the influence of image filtering techniques and methods for the evaluation of significance of pixels on the conversion process. Furthermore, it proposes several novel approaches for a compression of the Delaunay triangulation and compares them with existing solutions. [ABSTRACT FROM AUTHOR]

To make an encoder extremely simple by eliminating motion prediction/compensation from encoder, source coding with side information has been investigated based on the Wyner-Ziv theorem as the basic coding principle. However, the frame interpolation at decoder which is essential for redundancy elimination makes erroneous side information when the basic assumption of linear motion between frames is not satisfied. In this paper, we propose a new Wyner-Ziv video coding scheme featuring side matching in the frame interpolation to improve the side information. In the proposed scheme, Wyner-Ziv decoder compensates wrong blocks by side information using side matching and bi-directional searching. The noise reduction in side information allows the proposed algorithm to achieve coding improvements not only in bit-rate but also in PSNR. Results of our experiments show improvement of PSNR up to 0.4dB. [ABSTRACT FROM AUTHOR]

Wyner-Ziv video coding has become one of the hottest research topics in the video coding community due to the conceptual, theoretical and functional novelties it brings. Among the many practical architectures already available, feedback channel-based with channel coding, e.g. LDPC and turbo codes, solutions are rather popular. These solutions rely on decoder motion estimation based on periodic Intra coded key frames, setting the so-called GOP size, very much like in conventional video coding. This paper targets the rate-distortion and complexity performance study of this type of Wyner-Ziv coding solution as a function of the GOP size, considering both LPDC and turbo codes. [ABSTRACT FROM AUTHOR]

This paper investigates an extension of the SP picture from the H.264/AVC to the scalable video coding (SVC), which has been recently developed and standardized as the scalable extension of the H.264/AVC. In comparison with the scalable profiles of previous video coding standards, the SVC has achieved significant improvement in both coding efficiency and scalability in temporal, spatial and fidelity, which efficiently provides coded stream wide adaptivity to dynamic network conditions as well as diverse clients. In communication environments, this efficient adaptivity can be provided by bit stream switching between different scalable layers. The current SVC supports bit stream switching only at instantaneous decoding refresh (IDR) access unit. However, in order to provide instantaneous switching capability, the IDR picture needs to be frequently coded in the SVC stream, which dramatically decreases the coding efficiency. Therefore, SP picture for the SVC is proposed in this paper for efficient bit stream switching. Performance analysis shows that the SP picture for the SVC provides an average 1.2 dB PSNR enhancement over the IDR picture while providing similar functionalities. [ABSTRACT FROM AUTHOR]

This paper presents an application of the hidden Markov models (HMMs) to the recognition of snakes behaviors, an important and hard problem that, as far as the authors know, has not been tackled before, by the computer vision community. Experiments were conducted using different HMM configurations, including modifications on the number of internal states and the initialization procedures. The best results have shown a 84% correct classification rate, using HMMs with 4 states and an initialization procedure based on the K-Means algorithm. [ABSTRACT FROM AUTHOR]

A complete system for creating the performance of a virtual character is described. Stereo web-cameras perform marker based motion capture to obtain rigid head motion and non-rigid facial expression motion. Acquired 3D points are then mapped onto a 3D face model with a virtual muscle animation to create face expressions. Muscle inverse kinematics updates muscle contraction parameters based on marker motion to create the character's performance. Advantages of the system are reduced character creation time by using virtual muscles and a dynamic skin model, a novel way of applying markers to a face animation system, and its low cost hardware requirements, capable of running on standard hardware and making it suitable for interactive media in end-user environments. [ABSTRACT FROM AUTHOR]

This paper proposes the use of simulations as a design mechanism for visual inspection systems of specular surfaces. The system requirements and the characteristics of the objects involve a technological design problem for each of the solutions to be developed. A generic model is proposed. It may be adapted or particularised to solve specific inspection problems using simulations. The method results in a flexible low cost design, reducing the distance between the design model and system implementation in a manufacturing procedure. The proposed simulator generates model-based architectures. The paper shows the results on application of metallized automobile logos. [ABSTRACT FROM AUTHOR]

This paper presents a novel approach for automatic people counting in videos captured through a conventional closed-circuit television (CCTV) using computer vision techniques. The proposed approach consists of detecting and tracking moving objects in video scenes to further counting them when they enters into a virtual counting zone defined in the scene. One of the main problems of using conventional CCTV cameras is that they are usually not placed into a convenient position for counting and this may cause a lot of occlusions between persons when they are walking very close or in groups. To tackle this problem two strategies are investigated. The first one is based on two thresholds which are related to the average width and to the average area of a blob top zone, which represents a person head. By matching the width and the head region area of a current blob against these thresholds it is possible to estimate if the blob encloses one, two or three persons. The second strategy is based on a zoning scheme and extracts low level features from the top region of the blob, which is also related to a person head. Such feature vectors are used together with an instance-based classifier to estimate the number of persons enclosed by the blob. Experimental results on videos from two different databases have shown that the proposed approach is able to count the number of persons that pass through a counting zone with accuracy higher than 85%. [ABSTRACT FROM AUTHOR]

Recently, more and more computer vision researchers are paying attention to error analysis so as to fulfill various accuracy requirements arising from different applications. As a geometric invariant under projective transformations, cross-ratio is the basis of many recognition and reconstruction algorithms which are based on projective geometry. We propose an efficient way of analyzing localization error for computer vision systems which use cross-ratios in planar localization. By studying the inaccuracy associated with cross-ratio-based computations, we inspect the possibility of using linear transformation to approximate localization error due to 2-D noises of image extraction for reference points. Based on such a computationally efficient analysis, a practical way of choosing point features in an image so as to establish the probabilistically most accurate planar location system using cross-ratios is developed. [ABSTRACT FROM AUTHOR]

Research on image-based 3D reconstruction has recently shown a lot of good results, but it assumes precise target objects are already segmented from each input image. Traditionally, background subtraction was used to segment the target objects, but it can yield serious problems, such as noises and holes. To precisely segment the target objects, graph cuts have recently been used. Graph cuts showed good results in many engineering problems, as they can globally minimize energy functions composed of data terms and smooth terms, but it is difficult to automatically obtain prior information necessary for data terms. Depth information generated by stereo vision was used as prior information, which shows good results in their experiments, but it is difficult to calculate depth information for 3D face reconstruction, as the most of faces have homogeneous regions. In this paper, we propose better foreground segmentation method for 3D face reconstruction using graph cuts. The foreground objects are approximately segmented from each background image using background subtraction to assist to estimate data terms of energy functions, and noises and shadows are removed from the segmented objects to reduce errors of prior information. Removing the noises and shadows should cause to lose detail in the foreground silhouette, but smooth terms that assign high costs if neighboring pixels are not similar can fill out the lost silhouette. Consequently, the proposed method can segment more precise target objects by globally minimizing the energy function composed of smooth terms and approximately estimated data terms using graph cuts. [ABSTRACT FROM AUTHOR]

In this paper, we present human robot interaction techniques such as face pose and hand gesture for efficient viewing comics through the robot. For the controlling of the viewing order of the panel, we propose a robust face pose recognition method using the pose appearance manifold. We represent each pose of a person's face as connected low-dimensional appearance manifolds which are approximated by the affine plane. Then, face pose recognition is performed by computing the minimal distance from the given face image to the sub-pose manifold. To handle partially occluded faces, we generate an occlusion mask and then put the lower weights on the occluded pixels of the given image to recognize occluded face pose. For illumination variations in the face, we perform coarse normalization on skin regions using histogram equalization. To recognize hand gestures, we compute the center of gravity of the hand using skeleton algorithm and count the number of active fingers. Also, we detect index finger's moving direction. The contents in the panel are represented by the scene graph and can be updated according to the user's control. Based on the face pose and hand gesture recognition result, an audience can manipulate contents and finally appreciate the comics in his own style. [ABSTRACT FROM AUTHOR]

In this paper, an efficient and robust algorithm is proposed for image registration, where the images have been acquired at different times, by different sensors and some changes may take place in the scene during the time interval when the images were taken. By simulating human vision behaviors, image registration is carried out through a two-stage process. First, the rotation angles are computed by comparing the distributions of gradient orientations, which is implemented by a simple 1-D correlation. Then, a new approach is presented to detect significant corners in two images and the correspondences are established between corners in two images. At this time, the orientation difference has been removed between the images, so the corner correspondences can be established more efficiently. To account for the false corner correspondences, the voting method is used to determine the transformation parameters. Experimental results are also given. [ABSTRACT FROM AUTHOR]

A novel approach for the detection and classification of human movements in videos scenes is presented in this paper. It consists in detecting, segmenting and tracking foreground objects in video scenes to further classify their movements as conventional or non-conventional. From each tracked object in the scene, features such as position, speed, changes in direction and temporal consistency of the bounding box dimension are extracted. These features make up feature vectors that are stored together with labels that categorize the movement and which are assigned by human supervisors. At the classification step, an instance-based learning algorithm is used to classify the object movement as conventional or non-conventional. For this aim, feature vectors computed from objects in motion are matched against reference feature vectors previously labeled. Experimental results on video clips from two different databases (Parking Lot and CAVIAR) have shown that the proposed approach is able to detect non-conventional human movements in video scenes with accuracies between 77% and 82%. [ABSTRACT FROM AUTHOR]

The implementation of a specific image recognition technique for an artificial vision system is presented. The proposed algorithm involves two steps. First, smaller images are obtained using Discrete Wavelet Transform (DWT) after four stages of decomposition and taking only the approximations. This way the volume of information to process is reduced considerably and the system memory requirements are reduced as well. Another purpose of DWT is to filter noise that could be induced in the images. Second, the Morphological Associative Memories (MAM) are used to recognize landmarks. The proposed algorithm provides flexibility, possibility to parallelize algorithms and high overall performance of hardware implemented image retrieval system. The resulted hardware implementation has low memory requirements, needs in limited arithmetical precision and reduced number of simple operations. These benefits are guaranteed due to the simplicity of MAM learning/restoration process that uses simple morphological operations, dilation and erosion, in other words, MAM calculate maximums or minimums of sums. These features turn out the artificial vision system to be robust and optimal for the use in realtime autonomous systems. The proposed image recognition system has, among others, the following useful features: robustness to the noise induced in the patter to process, high processing speed, and it can be easily adapted to diverse operation circumstances. [ABSTRACT FROM AUTHOR]

Image registration based on landmarks and radial basis functions (e.g. thin plate splines) results in global changes and deformation spreads over the entire resampled image. This paper presents a radial basis function for registration of local changes. The proposed research was based on study/analysis of profile for different radial basis functions, supporting local changes. The proposed function was designed to overcome the weaknesses, observed in other radial basis functions. The results are analyzed/compared on the basis of different properties and parameters discussed in this paper. Experimental results show that the proposed function improves the registration accuracy. [ABSTRACT FROM AUTHOR]

The performance of Support Vector Machines, as many other machine learning algorithms, is very sensitive to parameter tuning, mainly in real world problems. In this paper, two well known and widely used SVM implementations, Weka SMO and LIBSVM, were compared using Simulated Annealing as a parameter tuner. This approach increased significantly the classification accuracy over the Weka SMO and LIBSVM standard configuration. The paper also presents an empirical evaluation of SVM against AdaBoost and MLP, for solving the leather defect classification problem. The results obtained are very promising in successfully discriminating leather defects, with the highest overall accuracy, of 99.59%, being achieved by LIBSVM tuned with Simulated Annealing. [ABSTRACT FROM AUTHOR]

This paper presents a vision-based method for tracking guitar fingerings played by guitar players from stereo cameras. We propose a novel framework for colored finger markers tracking by integrating a Bayesian classifier into particle filters, with the advantages of performing automatic track initialization and recovering from tracking failures in a dynamic background. ARTag (Augmented Reality Tag) is utilized to calculate the projection matrix as an online process which allow guitar to be moved while playing. By using online adaptation of color probabilities, it is also able to cope with illumination changes. [ABSTRACT FROM AUTHOR]

Motion extraction is an essential work in facial expression analysis because facial expression usually experiences rigid head rotation and non-rigid facial expression simultaneously. We developed a system to separate non-rigid motion from large rigid motion over an image sequence based on the incremental perspective motion model. Since the parameters of this motion model are able to not only represnt the global rigid motion but also localize the non-rigid motion, thus this motion model overcomes the limitations of existing methods, the affine model and the 8-parameter perspective projection model, in large head rotation angles. In addition, since the gradient descent approach is susceptible to local minimum during the motion parameter estimation process, a multi-resolution approach is applied to optimize initial values of parameters at the coarse level. Finally, the experimental result shows that our model has promising performance of separating non-rigid motion from rigid motion. [ABSTRACT FROM AUTHOR]

Significant achievements have been attained in the field of dense stereo correspondence by local algorithms based on an adaptive support. Given the problem of matching two correspondent pixels within a local stereo process, the basic idea is to consider as support for each pixel only those points which lay on the same disparity plane, rather than those belonging to a fixed support. This paper proposes a novel support aggregation strategy which includes information obtained from a segmentation process. Experimental results on the Middlebury dataset demonstrate that our approach is effective in improving the state of the art. [ABSTRACT FROM AUTHOR]

A human body measurement system using multiple viewpoints is proposed. Whole human body data taken from systems, which have been developed with a few viewpoints, have not been successful at acquiring due to occlusion. It is proposed that data can be successfully obtained by a method using multiple rangefinders correctly. Four compact rangefinders are installed in a pole. Those three pole units, with 12 rangefinders, are assigned around a person. Multiple viewpoint range images allow the 3D shape reconstruction of a human body. Then a morphable model is adapted to whole human body data. The measurement time is 2 seconds and the average error is found to be 1.88 mm. In this paper, the system configuration, calibration, morphable model and experimental results are described. [ABSTRACT FROM AUTHOR]

In recent years, there has been increased interest in characterizing and extracting 3D information from 2D images for human tracking and identification. In this paper, we propose a single view-based framework for robust estimation of height and position. In the proposed method, 2D features of target object is back-projected into the 3D scene space where its coordinate system is given by a rectangular marker. Then the position and the height are estimated in the 3D space. In addition, geometric error caused by inaccurate projective mapping is corrected by using geometric constraints provided by the marker. The accuracy and the robustness of our technique are verified on the experimental results of several real video sequences from outdoor environments. [ABSTRACT FROM AUTHOR]

This paper presents a dynamic heart model based on a parallelized space-time adaptive mesh refinement algorithm (AMRA). The spatial and temporal simulation method of the anisotropic excitable media has to achieve great performance in distributed processing environment. The accuracy and efficiency of the algorithm was tested for anisotropic and inhomogeneous 3D domains using ten Tusscher's and Nygen's cardiac cell models. During propagation of depolarization wave, the kinetic, compositional and rotational anisotrophy is included in the tissue, organ and torso model. The generated inverse ECG with conventional and parallelized algorithm has the same quality, but a speedup of factor 200 can be reached using AMRA modeling and single instruction multiple data (SIMD) programming of the video cards. These results suggest that a powerful personal computer will be able to perform a one-second long simulation of the spatial electrical dynamics of the heart in approximately five minutes. [ABSTRACT FROM AUTHOR]

In this article a robust and real-time hand gesture detection and recognition system for dynamic environments is proposed. The system is based on the use of boosted classifiers for the detection of hands and the recognition of gestures, together with the use of skin segmentation and hand tracking procedures. The main novelty of the proposed approach is the use of innovative training techniques - active learning and bootstrap -, which allow obtaining a much better performance than similar boosting-based systems, in terms of detection rate, number of false positives and processing time. In addition, the robustness of the system is increased due to the use of an adaptive skin model, a color-based hand tracking, and a multi-gesture classification tree. The system performance is validated in real video sequences. [ABSTRACT FROM AUTHOR]

Magnetic Resonance Imaging (MRI) has become an important tool for dynamic clinical studies. Regrettably, the long acquisition time is still a challenge in dynamic MRI. Several undersampled reconstruction techniques have been developed to speed up the acquisition without significantly compromising image quality. Most of these methods are based on modeling the pixel intensity changes. Recently, we introduced a new approach based on the motion estimation of each object element (obel, a piece of tissue). Although the method works well, the outcome is a trade off between the maximum undersampling factor and the motion estimation accuracy. In this work we propose to improve its performance through the use of additional data from multiple coils acquisition. Preliminary results on cardiac MRI show that further undersampling and/or improved reconstruction accuracy is achieved using this technique. Furthermore, an approximation of the vector field of motion is obtained. This method is appropriate for sequences where the obels' intensity through time is nearly constant. [ABSTRACT FROM AUTHOR]

This paper presents a new algorithm for spatially modulated tone mapping in Standard Dynamic Range (SDR) images. The method performs image enhancement by lightening the tones in the under-exposured regions while darkening the tones in the over-exposured, without affecting the correctly exposured ones. The tone mapping function is inspired by the shunting characteristics of the center-surround cells of the Human Visual System (HVS). This function is modulated differently for every pixel, according to its surround. The surround is calculated using a new approach, based on the oriented cells of the HVS, which allows it to adapt its shape to the local contents of the image and, thus, minimize the halo effects. The method has low complexity and can render 1MPixel images in approximately 1 second when executed by a conventional PC. [ABSTRACT FROM AUTHOR]

In this paper, we present an external calibration technique for typical multi-camera system. The technique is very handy in practice using a simple planar pattern. Based on homography, an efficient pair-wise estimation method is proposed to recover the rigid rotation and translation between neighboring cameras. By registering all these partial calibrated structures, complete calibration of the multi-camera system is accomplished. Experiments with both simulated and real data show that accurate and stable calibration results can be achieved by the proposed method. [ABSTRACT FROM AUTHOR]

This paper considers the problem of Bayesian inference in dynamical models with time-varying dimension. These models have been studied in the context of multiple target tracking problems and for estimating the number of components in mixture models. Traditional solutions for the single target tracking problem becomes infeasible when the number of targets grows. Furthermore, when the number of targets is unknown and the number of observations is influenced by misdetections and clutter, then the problem is complex. In this paper, we consider a marked Poisson process for modeling the time-varying dimension problem. Another solution which has been proposed for this problem is the Probability Hypothesis Density (PHD) filter, which uses a random set formalism for representing the time-varying nature of the state and observation vectors. An important feature of the PHD and the proposed method is the ability to perform sensor data fusion by integrating the information from the multiple observations without an explicit data association step. However, the method proposed here differs from the PHD filter in that uses a Poisson point process formalism with discretized spatial intensity. The method can be implemented with techniques similar to the standard particle filter, but without the need for specifying birth and death probabilities for each target in the update and filtering equations. We show an example based on ultrasound acoustics, where the method is able to represent the physical characteristics of the problem domain. [ABSTRACT FROM AUTHOR]

Homographies are widely used in tasks like camera calibration, tracking, mosaicing or motion estimation and numerous linear and non linear methods for homography estimation have been proposed in the case of classical cameras. Recently, some works have also proved the validity of homography for catadioptric cameras but only a linear estimator has been proposed. In order to improve the estimation based on correspondence features, we suggest in this article some non linear estimators for catadioptric sensors. Catadioptric camera motion estimation from a sequence of a planar scene is the proposed application for the evaluation and the comparison of these estimation methods. Experimental results with simulated and real sequences show that non linear methods are more accurate. [ABSTRACT FROM AUTHOR]

This study develops a 3D facial reconstruction system, which consists of five modules, using the orthographic projection model to approximate the perspective projection model. The first module identifies a number of feature points on the face and tracks these feature points over a sequence of facial images by the optical flow technique. The second module applies the factorization method to the orthographic model to reconstruct a 3D human face. The facial images are acquired using a pinhole camera, which are based on a perspective projection model. However, the face is reconstructed using an orthographic projection model. To compensate for the difference between these two models, the third module implements a simple and efficient method for approximating the perspective projection model. The fourth module overcomes the missing point problem, commonly arising in 3D reconstruction applications. Finally, the fifth module implements a smoothing process for the 3D surface by interpolating additional vertices. [ABSTRACT FROM AUTHOR]

3D data is getting popular which offers more details and accurate information for posture recognition. However, it leads to computational hurdles and is not suitable for real time application. Therefore, we introduce a dimension reduction method using meshless parameterization with cylindrical virtual boundary for 3D posture representation. The meshless parameterization is based on convex combination approach which has good properties, such as fast computation and one-to-one mapping characteristic. This method depends on the number of boundary points. However, 3D posture reconstruction using silhouettes extraction from multiple cameras had resulted various number of boundary points. Therefore, a cylindrical virtual boundary points is introduced to overcome the inconsistency of 3D reconstruction boundary points. The proposed method generates five slices of 2D parametric appearance to represent a 3D posture for recognition purpose. [ABSTRACT FROM AUTHOR]

This paper presents a new online Video-Based Rendering method that creates simultaneously multiple views at every new frame. Our system is especially designed for communication between mobile phones using autostereoscopic display and computers. The new views are computed from 4 webcams connected to a computer and are compressed in order to be transferred to the mobile phone. Thanks to GPU programming, our method provides up to 16 images of the scene in real-time. The use of both GPU and CPU makes our method work on only one consumer grade computer. [ABSTRACT FROM AUTHOR]

The purpose of estimating subjective image quality is to provide best-quality image content to users in diverse fields. Image quality is subjective, therefore very difficult to estimate accurately, and so many researchers have proposed psychophysical experimentation as a means of estimating it. Conventional methods describe the relationship between the subjective preference and the perceived contrast as an "inverted U" shape. However, the relationship was resulted from only a few, high-quality images. Thus, they are inadequate for general image-quality estimation. In this paper, we carried out two experiments using a dataset with untransformed and various images. We discovered an important property that the preference increases in proportion to the perceived contrast. The result shows us not only that our experimentation can reduce MSE of image-quality estimation by approximately 40% over the previous methods, but also that it can be applied in various applications. [ABSTRACT FROM AUTHOR]

This paper describes a novel real-time technique to determine horizontal human face pose from a video color sequence. The idea underlying this technique is that when head is at an arbitrary pose to the right or left, there are significant relationships between the distance from center of both pupils to head center, and the distance between both pupils. From these distances, we compute a ratio known as "horizontal ratio". This ratio, besides being advantageous in the sense that it reduces the dependency on facial features tracking accuracy and robust to noise, is actually the quantity that is used to determine the horizontal human face pose. The technique is simple, computational cheap and requires only information that is usually retrievable from a face and facial feature tracker. [ABSTRACT FROM AUTHOR]

Most commercial digital cameras use a single electronic sensor overlaid with a color filter array (CFA) to capture imagery. Since only one primary color is sampled in each pixel, the missing color primaries must be reconstructed by interpolation. In this paper, an adaptive demosaicking scheme for CFA interpolation is proposed. The scheme uses intra-channel correlation, color difference correlation, constant hue, and luminance-color difference correlation is proposed. A rough interpolation is first implemented by bilinear interpolation. Then the color difference correlation and constant hue are successively used to update the missing color primaries. To obtain high quality color images, an adaptive algorithm using luminance-color difference correlation and the information of edge direction is iteratively applied to improve the image quality around the edges. Simulation results demonstrate that the image quality of the proposed algorithm is better than that of the approach using color difference correlation in terms of peak signal-to-noise ratio (PSNR). [ABSTRACT FROM AUTHOR]

Automated inspection using multiple views (AMVI) has been recently developed to automatically detect flaws in manufactured objects. The principal idea of this strategy is that, unlike the noise that appears randomly in images, only the flaws remain stable in a sequence of images because they remain in their position relative to the movement of the object being analyzed. This investi- gation proposes a new strategy, based on the detection of flaws in a non- calibrated sequence of images. The method uses a scheme of elimination of potential flaws in two and three views. To improve the performance, intermediate blocks are introduced that eliminate those hypothetical flaws that are regular regions and real flaws. Use is made of images captured in a non-calibrated vision system, so there are no optical, geometric and noise disturbances in the image, forcing the proposed method to be robust, so that it can be applied in industry as a quality control method in non-calibrated vision systems. the results show that it is possible to detect the real flaws and at the same time decrease most of the false alarms. [ABSTRACT FROM AUTHOR]

Let p and q be two points on the surface of a polytope Π. This paper provides a rubberband algorithm for computing a Euclidean shortest path between p and q (a so-called surface ESP) that is contained on the surface of Π. The algorithm has $\kappa_1(\varepsilon) \cdot \kappa_2(\varepsilon) \cdot {\cal O}(n^2)$ time complexity, where n is the number of vertices of Π, κi(ε) = (L0i − Li)/ε, for the true length Li of some shortest path with initial (polygonal path) length L0i (used when approximating this shortest path), for i = 1, 2. Rubberband algorithms follow a straightforward design strategy, and the proposed algorithm is easy to implement and thus of importance for applications, for example, when analyzing 3D objects in 3D image analysis, such as in biomedical or industrial image analysis, using 3D image scanners. [ABSTRACT FROM AUTHOR]

A new and efficient method for combining iris biometrics with custom cryptographic schemes to obtain an efficient biocryptosystem is proposed in this paper. Though the method structure is basically derived from a previously described biocryptosystem scheme, the introduction of new image processing methods alongside with efficient utilization of traditional methods show promising developments compared with the previous biocryposystem especially in the field of generating longer cryptographic key strings while keeping the system quality [ABSTRACT FROM AUTHOR]