Existing deep learning-based interactive image segmentation approaches typically assume the target-of-interest is always a single object and fail to account for the potential diversity in user expectations, thus requiring excessive user input when it comes to segmenting an object part or a group of objects instead. Motivated by the observation that the object part, full object, and a collection of objects essentially differ in size, we propose a new concept called scale-diversity, which characterizes the spectrum of segmentations w.r.t. different scales. To address this, we present MultiSeg, a scale-diverse interactive image segmentation network that incorporates a set of two-dimensional scale priors into the model to generate a set of scale-varying proposals that conform to the user input. We explicitly encourage segmentation diversity during training by synthesizing diverse training samples for a given image. As a result, our method allows the user to quickly locate the closest segmentation target for further refinement if necessary. Despite its simplicity, experimental results demonstrate that our proposed model is capable of quickly producing diverse yet plausible segmentation outputs, reducing the user interaction required, especially in cases where many types of segmentations (object parts or groups) are expected.

In this work, we propose a hierarchical Bayesian network based point set registration method to solve missing correspondences and various massive outliers. We construct this network first using the finite Student s t latent mixture model (TLMM), in which distributions of latent variables are estimated by a tree-structured variational inference (VI) so that to obtain a tighter lower bound under the Bayesian framework. We then divide the TLMM into two different mixtures with isotropic and anisotropic covariances for correspondences recovering and outliers identification, respectively. Finally, the parameters of mixing proportion and covariances are both taken as latent variables, which benefits explaining of missing correspondences and heteroscedastic outliers. In addition, a cooling schedule is adopted to anneal prior on covariances and scale variables within designed two phases of transformation, it anneal priors on global and local variables to perform a coarse-to- fine registration. In experiments, our method outperforms five state-of-the-art methods in synthetic point set and realistic imaging registrations.

Deep learning has been applied to segment buildings from high-resolution images with promising results. However, there still exist the problems stemming from training on split patches and class imbalances. To overcome these problems, we propose a dual-resolution U-Net that uses pairs of images as inputs to capture both high and low resolution features. We also employ a soft Jaccard loss to place more emphasis on the sparse and low accuracy samples. The images from different regions are further balanced according to their building densities. With our architecture, we achieved state-of-the-art results on the Inria aerial image labeling dataset without any post-processing.

The interactive image segmentation model allows users to iteratively add new inputs for refinement until a satisfactory result is finally obtained. Therefore, an ideal interactive segmentation model should learn to capture the user's intention with minimal interaction. However, existing models fail to fully utilize the valuable user input information in the segmentation refinement process and thus offer an unsatisfactory user experience. In order to fully exploit the user-provided information, we propose a new deep framework, called Regional Interactive Segmentation Network (RIS-Net), to expand the field-of-view of the given inputs to capture the local regional information surrounding them for local refinement. Additionally, RIS-Net adopts multiscale global contextual information to augment each local region for improving feature representation. We also introduce click discount factors to develop a novel optimization strategy for more effective end-to-end training. Comprehensive evaluations on four challenging datasets well demonstrate the superiority of the proposed RIS-Net over other state-of-the-art approaches.

We present a new point set registration method with global-local correspondence and transformation estimation (GL-CATE). The geometric structures of point sets are exploited by combining the global feature, the point-to-point Euclidean distance, with the local feature, the shape distance (SD) which is based on the histograms generated by an elliptical Gaussian soft count strategy. By using a bidirectional deterministic annealing scheme to directly control the searching ranges of the two features, the mixture-feature Gaussian mixture model (MGMM) is constructed to recover the correspondences of point sets. A new vector based structure constraint term is formulated to regularize the transformation. The accuracy of transformation updating is improved by constraining spatial structure at both global and local scales. An annealing scheme is applied to progressively decrease the strength of the regularization and to achieve the maximum overlap. Both of the aforementioned processes are incorporated in the EM algorithm, a unified optimization framework. We test the performances of our GL-CATE in contour registration, sequence images, real images, medical images, fingerprint images and remote sensing images, and compare with eight state-of-the-art methods where our GL-CATE shows favorable performances in most scenarios.

Region-based active contour models are effective in segmenting images with poorly defined boundaries but often fail when applied to images containing intensity inhomogeneity. The traditional models utilize pixel intensity and are very sensitive to parameter tuning. On the other hand, machine learning algorithms are highly effective in handling inhomogeneities but often result in noise from misclassified pixels. In addition, there is no objective function. We propose a framework which integrates machine learning with a region-based active contour model. Classification probability scores from machine learning algorithm, which are regularized using a non-linear function, are used to replace the pixel intensity values during energy minimization. In our experiments, we integrate the k-nearest neighbours and the support vector machine with the Chan-Vese method and compare the results obtained with the traditional methods of Chan-Vese and Li et al. The proposed framework gives better accuracy and less sensitive to parameter tuning.

Brain tumor segmentation, an essential but challenging task, has long attracted much attention from the medical imaging community. Recently, successful applications of sparse coding and dictionary learning has emerged in various vision problems including image segmentation. In this paper, a superpixel-based framework for automated brain tumor segmentation is introduced. The kernel trick is adopted in dictionary learning to transform superpixel-level features to a high-dimensional feature space where their nonlinear similarities are considered to generate discriminative sparse codes. A graph is constructed from the approximation errors given by dictionaries modeling different brain tumor structures so that superpixels belonging to particular tumor regions can be efficiently identified. The proposed framework is evaluated on brain magnetic resonance images of high-grade glioma (HGG) patients provided by the multi-modal Brain Tumor Segmentation (BRATS) Benchmark. Results show that the proposed framework achieves competitive performance when compared with the state-of-the-art methods.

Edge-based active contour models are effective in segmenting images with intensity inhomogeneity but often fail when applied to images containing poorly defined boundaries, such as in medical images. Traditional edge-stop functions (ESFs) utilize only gradient information, which fails to stop contour evolution at such boundaries because of the small gradient magnitudes. To address this problem, we propose a framework to construct a group of ESFs for edge-based active contour models to segment objects with poorly defined boundaries. In our framework, which incorporates gradient information as well as probability scores from a standard classifier, the ESF can be constructed from any classification algorithm and applied to any edge-based model using a level set method. Experiments on medical images using the distance regularized level set for edge-based active contour models as well as the k-nearest neighbours and the support vector machine confirm the effectiveness of the proposed approach.

Abstract In volume visualization, transfer functions are used to classify the volumetric data and assign optical properties to the voxels. In general, transfer functions are generated in a transfer function space, which is the feature space constructed by data values and properties derived from the data. If volumetric objects have the same or overlapping data values, it would be difficult to separate them in the transfer function space. In this paper, we present a rule‐enhanced transfer function design method that allows important structures of the volume to be more effectively separated and highlighted. We define a set of rules based on the local frequency distribution of volume attributes. A rule‐selection method based on a genetic algorithm is proposed to learn the set of rules that can distinguish the user‐specified target tissue from other tissues. In the rendering stage, voxels satisfying these rules are rendered with higher opacities in order to highlight the target tissue. The proposed method was tested on various volumetric datasets to enhance the visualization of important structures that are difficult to be visualized by traditional transfer function design methods. The results demonstrate the effectiveness of the proposed method.

The surface bi-directional reflectance distribution function (BRDF) can be used to distinguish different materials. The BRDFs of many real materials are near isotropic and can be approximated well by a 2D function. When the camera principal axis is coincident with the surface normal of the material sample, the captured BRDF slice is nearly 1D, which suffers from significant information loss. Thus, improvement in classification performance can be achieved by simply setting the camera at a slanted view to capture a larger portion of the BRDF domain. We further use a handheld flashlight camera to capture a 1D BRDF slice for material classification. This 1D slice captures important reflectance properties such as specular reflection and retro-reflectance. We apply these results on ink classification, which can be used in forensics and analyzing historical manuscripts. For the first time, we show that most of the inks on the market can be well distinguished by their reflectance properties.

The non-stationarity inherent across sessions recorded on different days poses a major challenge for practical electroencephalography (EEG)-based Brain Computer Interface (BCI) systems. To address this issue, the computational model trained using the training data needs to adapt to the data from the test sessions. In this paper, we propose a novel approach to compute the variations between labelled training data and a batch of unlabelled test data based on the geodesic-distance of the discriminative subspaces of EEG data on the Grassmann manifold. Subsequently, spatial filters can be updated and features that are invariant against such variations can be obtained using a subset of training data that is closer to the test data. Experimental results show that the proposed adaptation method yielded improvements in classification performance.

This paper addresses the key issue of discriminative feature extraction of electroencephalogram (EEG) signals in brain-computer interfaces. Recent advances in neuroscience indicate that multiple brain regions can be activated during motor imagery. The signal propagation among the regions can give rise to spurious effects in identifying event-related desynchronization/synchronization for discriminative motor imagery detection in conventional feature extraction methods. Particularly, we propose that computational models which account for both signal propagation and volume conduction effects of the source neuronal activities can more accurately describe EEG during the specific brain activities and lead to more effective feature extraction. To this end, we devise a unified model for joint learning of signal propagation and spatial patterns. The preliminary results obtained with real-world motor imagery EEG data sets confirm that the new methodology can improve classification accuracy with statistical significance.

While most photometric stereo algorithms assume Lambertian reflectance model for its simplicity, surfaces in general are not Lambertian. Some recent photometric stereo algorithms [1, 2] can recover iso-depth and iso-grandient contours for arbitrary isotropic surfaces by exploiting more general reflectance properties. However, further reconstruction of the 3D surface requires additional information, such as the normal at a point or depth along a boundary as suggested in [2]. This paper addresses the problem of reconstructing the surface from iso-depth and iso-gradient contours without such kind of additional information. The experiment results demonstrate the effectiveness of our algorithm.

This article addresses the broad question of understanding whether and how a combination of tactile and visual information could be used to enhance the experience of music by the hearing impaired. Initially, a background survey was conducted with hearing-impaired people to find out the techniques they used to “listen” to music and how their listening experience might be enhanced. Information obtained from this survey and feedback received from two profoundly deaf musicians were used to guide the initial concept of exploring haptic and visual channels to augment a musical experience. The proposed solution consisted of a vibrating “Haptic Chair” and a computer display of informative visual effects. The Haptic Chair provided sensory input of vibrations via touch by amplifying vibrations produced by music. The visual display transcoded sequences of information about a piece of music into various visual sequences in real time. These visual sequences initially consisted of abstract animations corresponding to specific features of music such as beat, note onset, tonal context, and so forth. In addition, because most people with impaired hearing place emphasis on lip reading and body gestures to help understand speech and other social interactions, their experiences were explored when they were exposed to human gestures corresponding to musical input. Rigorous user studies with hearing-impaired participants suggested that musical representation for the hearing impaired should focus on staying as close to the original as possible and is best accompanied by conveying the physics of the representation via an alternate channel of perception. All the hearing-impaired users preferred either the Haptic Chair alone or the Haptic Chair accompanied by a visual display. These results were further strengthened by the fact that user satisfaction was maintained even after continuous use of the system over a period of 3 weeks. One of the comments received from a profoundly deaf user when the Haptic Chair was no longer available (“I am going to be deaf again”), poignantly expressed the level of impact it had made. The system described in this article has the potential to be a valuable aid in speech therapy, and a user study is being carried out to explore the effectiveness of the Haptic Chair for this purpose. It is also expected that the concepts presented in this paper would be useful in converting other types of environmental sounds into a visual display and/or a tactile input device that might, for example, enable a deaf person to hear a doorbell ring, footsteps approaching from behind, or a person calling him or her, or to make understanding conversations or watching television less stressful. Moreover, the prototype system could be used as an aid in learning to play a musical instrument or to sing in tune. This research work has shown considerable potential in using existing technology to significantly change the way the deaf community experiences music. We believe the findings presented here will add to the knowledge base of researchers in the field of human–computer interaction interested in developing systems for the hearing impaired.

Effective learning and recovery of relevant source brain activity patterns is a major challenge to brain-computer interface using scalp EEG. Various spatial filtering solutions have been developed. Most current methods estimate an instantaneous demixing with the assumption of uncorrelatedness of the source signals. However, recent evidence in neuroscience suggests that multiple brain regions cooperate, especially during motor imagery, a major modality of brain activity for brain-computer interface. In this sense, methods that assume uncorrelatedness of the sources become inaccurate. Therefore, we are promoting a new methodology that considers both volume conduction effect and signal propagation between multiple brain regions. Specifically, we propose a novel discriminative algorithm for joint learning of propagation and spatial pattern with an iterative optimization solution. To validate the new methodology, we conduct experiments involving 16 healthy subjects and perform numerical analysis of the proposed algorithm for EEG classification in motor imagery brain-computer interface. Results from extensive analysis validate the effectiveness of the new methodology with high statistical significance.

Current techniques for investigating the functional roles of masticatory muscles are not suited to explaining subject-specific biomechanical relationship between mandibular movements and masticatory muscle activities. The aims of this study were to estimate the muscle tensions of subject-specific masticatory muscles in dental occlusion through the three-dimensional morphologic changes (3DMCs) of the muscles, and to explain the subject-specific biomechanical relationship between the mandibular movement and the muscle tensions. One healthy adult subject underwent magnetic resonance (MR) scans of the head at mandibular rest position (M0) and maximum intercuspation (M1). Based on the two sets of MR images, the mandibular movement was measured by the position changes of the mental protuberance of the mandible and the muscle tension from M0 to M1 for each masticatory muscle was estimated by its 3DMCs. The results showed the subject-specific biomechanical relationship between the mandibular movement and the muscle tensions, and the mandibular movement could be explained by these related muscle tensions anatomically and functionally.

The aims of this study were to present a method quantifying and visualizing the deformation of subject-specific lateral pterygoid muscles (LPM) during a simulated jaw-opening movement. A normal adult male subject underwent magnetic resonance (MR) scans of the head at three mandibular positions: mandibular rest (M0), medium jaw-opened (M1), and maximum jaw-opened (M2) positions. The 3D models of the LPM were reconstructed from the three sets of MR images. The deformations of each muscle in the two cases (M0->M1 and M1->M2) were quantified in terms of the displacements of region correspondences between the muscle models before and after the mandibular position changed. The 3D models of the subject-specific LPM were reconstructed, and the directions and magnitudes of deformations of each muscle in the two cases were accurately quantified and visualized in the three anatomic planes. The functional activities along the entire body and at specific compartments of subject-specific LPM were quantified and visualised using the quantified 3D deformations of the LPM as a new descriptor. The presented method defined the deformations of the subject-specific LPM, and revealed the anatomic architectural and biomechanical characteristics of the subject-specific LPM appropriately and meaningfully in the simulated jaw-opening movement.

Despite the popularity of spin images in surface matching and registration, disadvantages such as noise sensitivity and low discriminative ability still hindered their usefulness in real applications. In this paper, a novel approach was proposed for improving the spin images. The proposed method modified the standard spin images by using angle information between the normals of reference point and neighboring points. This information largely increased the robustness to noise without losing the intrinsic advantages of spin images. Moreover, signs were defined to incorporate the directions of angles which were shown to be able to further improve the descriptive power. Experiments were also conducted to show the outperformance of improved spin images under different levels of noise, and good agreements were obtained by comparing with the standard spin images and a recent popular 3D descriptor.

In this paper, we explore the use of spatial and frequency information of features in the biologically inspired model of HMAX. We discuss and refine previous models which use a similar framework and build specialized features which are better tuned to image structures by using unsupervised methods of clustering and picking the most frequent features using the statistics of the occurrence of the features in different spatial zones. Our classification results on the Caltech 101 dataset show significant improvements of up to 6% compared to previous improvements of the biologically inspired model of HMAX.

Phase unwrapping is an important and challenging problem in phase contrast magnetic resonance imaging (PC-MRI). In this paper, we propose a new algorithm for phase unwrapping based on graph cuts. Our algorithm takes a patch-based approach which has the advantages of simplicity and robustness to phase noise. To make use of temporal information from the neighboring frames as well as spatial information from the current frame, the energy function is designed to combine both spatial and temporal constraints. The proposed method has been tested with real PC-MRI data. Experimental results demonstrate that our algorithm is capable of unwrapping images with severe phase wrapping, and that it outperforms other existing popular phase unwrapping algorithms both qualitatively and quantitatively.

Blastocystis is a unicellular but polymorphic protozoan parasite causing digestive diseases in humans. Autophagy, a self-degradation process, is only recently found in Blastocystis. Identifying and enumerating autophagic Blastocystis cells using fluorescent microscopy are important in biology. Doing this manually is laborious and error-prone. This paper proposes image analysis techniques to automate the process. The difficulties are poor image quality and large variations in illumination and cell morphology. We divide the cells into several sub-classes of different morphology. Support vector machines are used to learn domain knowledge and classify the cells. Validation experiments on separate data sets show reliable performance for manually segmented cells with sensitivity 82.2% and specificity 86.7%. For automatically segmented cells, the sensitivity is the same. However, the specificity drops down to 68.4%. To our knowledge, this is the first attempt in automatic processing these images.

Computer generation of cell images can provide annotated data to simulate various imaging conditions with controllable parameters. Synthesized images based on simple models cannot reflect the complicated parameter constraints in simulating real objects in terms of their deformation with appropriate probabilities. Learning-based techniques can provide insight to these properties and impose constraints on deformation selections. In this work, we discuss the simulation of gray level images of healthy red blood cell populations. Different from existing techniques, we learn the unbiased average shape and deformation models of the cells. Both models are used to guide the selection of possible deformations. We also learn cell color models to govern the texture generation of simulated cells. We apply this technique to simulate cell populations and validate the results using cell segmentation and counting algorithms. The proposed learning and simulation technique is generic and can be applied to other types of cells as well.

Decomposition of cells in clumps is a difficult segmentation task requiring region splitting techniques. Techniques that do not employ prior shape constraints usually fail to achieve accurate segmentation. Those using shape constraints are unable to cope with large clumps and occlusions. In this work, we propose a model-constrained region splitting algorithm for cell clump decomposition. We build the cell model using joint probability distribution of invariant shape features. The shape model, the contour smoothness and the gradient information along the cut are used to optimize the splitting in a recursive manner. The short cut rule is also adopted as a strategy to speed up the process. The algorithm performs well in validation experiments using 60 images with 4516 cells and 520 clumps.

Baseline calibration is needed in most of SAR interferometry processing. An iterative optimization of baseline with constrain of relative platform position is presented in this paper. The SAR passes which gives inaccurate platform position is successfully detected and calibrated using this algorithm. After processing, new estimated baseline improves the quality of interferogram. Existence of reference Digital Elevation Model (DEM) error and atmospheric phase screen (APS) can also be detected from the convergence value. This method is based on a reversed concept of platform position estimation from interferometric result. Validation of method performed on multiple SAR images over Singapore.

Although dual-energy X-ray absorptiometry (DXA) offers an effective measurement of bone mineral density, it only provides a 2-D projected measurement of the bone mineral density. Clinical computed tomography (CT) imaging will have to be employed for measurement of 3-D bone mineral density. The typical dual energy process requires precise measurement of the beam spectral intensity at the 80 kVp and 120 kVp settings. However, this is not used clinically because of the extra radiation dosage and sophisticated hardware setup. We propose an accurate and fast approach to measure bone material properties with single energy scans. Beam hardening artifacts are eliminated by incorporating the polychromatic characteristics of the X-ray beam into the reconstruction process. Bone mineral measurement from single energy CT correction is compared with that of dual energy correction and the commonly used DXA. Experimental results show that single energy correction is compatible with dual energy CT correction in eliminating beam hardening artifacts and producing an accurate measurement of bone mineral density. We can then estimate Young's modulus, yield stress, yield strain and ultimate tensile stress of the bone, which are important data for patient specific therapy planning.

Music is a multi-dimensional experience informed by much more than hearing alone, and is thus accessible to people of all hearing abilities. In this paper we describe a prototype system designed to enrich the experience of music for the deaf by enhancing sensory input of information via channels other than in-air audio reception by the ear. The system has two main components-a vibrating 'Haptic Chair' and a computer display of informative visual effects that correspond to features of the music. The Haptic Chair provides sensory input of vibrations via touch. This system was developed based on an initial concept guided by information obtained from a background survey conducted with deaf people from multi-ethnic backgrounds and feedback received from two profoundly deaf musicians. A formal user study with 43 deaf participants suggested that the prototype system enhances the musical experience of a deaf person. All of the users preferred either the Haptic Chair alone (54%) or the Haptic Chair with the visual display (46%). The prototype system, especially the Haptic Chair was so enthusiastically received by our subjects that it is possible this system might significantly change the way the deaf community experiences music.

The construction of probabilistic liver atlases has received little attention in the past. Existing methods are based on landmarks and are sensitive to their choices and placements. We propose an iterative landmark-free method based on dense volumes to construct linear unbiased diffeomorphic probabilistic atlases from liver CT images. The linear averaging of the transformed images is set as the common target space followed by pairwise diffeomorphic registrations to warp all images to the target using a recent-proposed efficient deformation approach during each iteration cycle. Iterative pairwise registrations are directly used to handle possible large deformations without the need for an extra step to remove global deformations such as the use of affine transformations in traditional methods. Compared with those approaches estimating the unbiased atlas and the transformations groupwise simultaneously, the current method is more efficient. The efficiency and the convergence of our method have been demonstrated experimentally by validation using 25 CT liver sets.

In this work, video segmentation is viewed as an efficient intra-frame grouping temporally reinforced by a strong inter-frame coherence. Traditional approaches simply regard pixel motions as another prior in the MRF-MAP framework. Since pixel pre-grouping is inefficiently performed on every frame, the strong correlation between inter-frame groupings is largely underutilized. We exploit the inter-frame correlation to propagate trustworthy groupings from the previous frame. A preceding graph is constructed and labeled for the previous frame. It is temporally propagated to the current frame and validated by similarity measures. All unlabeled subgraphs are spatially aggregated for the final grouping. Experimental results show that the proposed approach is highly efficient for spatio-temporal segmentation. It makes good use of temporal correlation and produces satisfactory grouping results.

Automatic classification of working areas in peripheral blood smears can provide objective and reproducible quality control for the evaluation of smears and smear maker devices. However, it has drawn little research attention. In this paper we study this topic using image analysis and statistical pattern recognition methods. We employ generic features without requiring the extraction of individual cells. Two new spatial distribution features across scales are defined and utilized to classify working areas. We demonstrate that the only feature and method proposed in a similar work by others is insufficient to characterize the goodness of working areas, particularly the cell distribution. However, by utilizing it together with the features developed in this paper, we can achieve much better results. Our method has been tested on about 150 labeled images acquired from three malaria-infected Giemsa-stained blood smears using an oil immersion 100x objective lens.

This paper addresses the problem of extracting the road region in different driving environments with dynamic lighting changes. Previous approaches using Gaussian mixture models (GMM) have fixed number of models constructed from sample color data and could not keep models associated with shadows. As a result, although they work in some specific environments, they fail in other environments or in scenes with shadows. In this paper, we propose a new vision-based approach where flexible number of models are built from sample data. Those color samples are reliably collected from stereo-verified ground patches inside a pre-defined trapezoidal learning region. After model construction, models associated with shadows and highlights are detected and maintained. The advantages of this approach with respect to other techniques are that it gives more robust results and, in particular, recognizes shadows on road as drivable road surface instead of non-road.

This paper addresses the problem of extracting the road region in different driving environments with dynamic lighting changes, for driver-assistance applications. In this paper, we propose a stereo visual sensor system and a vision-based road extraction method in a new color space. The color space is designed such that it is representative of intrinsic reflectance of the road surface, and independent of illumination source. Our basic model of road color is a mixture of Gaussians in that color space, constructed from road sample pixels. Those color samples are reliably collected from stereo-verified ground patches inside a pre-defined trapezoidal learning region. The advantages of this system with respect to other systems are that it is more economical for driver-assistance applications while giving robust results and, in particular, recognizing shadows on road as drivable road surface instead of non-road.

The identification of chromosome abnormalities is an essential part of diagnosis and treatment of genetic disorders such as chromosomal syndromes and many types of cancer. Modern cytogenetic imaging techniques have improved the study of chromosome aberrations but they are most often used as adjuncts to traditional G-banded karyotype analysis. Molecular cytogenetic techniques such as comparative genomic hybridization, multicolor fluorescenceinsituhybridization, and spectral karyotyping (SKY) are able to detect chromosome copy number changes and complex structural aberrations in cancers, particularly in hematological malignancies and solid tumors. However, banded chromosome analysis is essential to distinguish between normal and abnormal chromosomes. Currently available cytogenetic imaging software is designed to classify only normal chromosomes. The identification of the banded regions involved in the abnormal chromosomes is done manually. In this paper, we propose an algorithm to automate the banding analysis of abnormal chromosomes by comparing the information obtained by SKY for precise identification of translocation break points. Our algorithm is based on the dynamic time warping method in order to overcome the problems due to the nonrigid nature of chromosomes. The method has been implemented and successfully applied to detect chromosome translocations, deletions, and duplications in cell lines derived from solid tumors.

Finding the shortest path from a given source to a given destination is a well known and widely applicable problem. Most of the work done in the area have used static route planning algorithms such as A*, Dijkstra’s, Bellman-Ford algorithm etc. Although these algorithms are said to be optimum, they are not capable of dealing with certain real life scenarios. For example, most of these single objective optimizations fails to find the equally good solutions when there is more than one optimum (shortest distance path, least congested path). We believe that the Genetic Algorithm (GA) based route planning algorithm proposed in this paper has the ability to tackle the above problems. In this paper, the proposed GA based route planning algorithm is successfully tested on the entire Singapore map with more than 10,000 nodes. Performance of the proposed GA is compared with an ant based path planning algorithm. Simulation results demonstrate the effectiveness of the proposed algorithm over ant based algorithm. Moreover, the proposed GA may be used as a basis for developing an intelligent route planning system.

Tracking objects using multiple sensors is more efficient than those using one sensor. In this paper, we proposed a method to fuse data from multiple sensors in Gaussian mixture probability hypothesis density filter. This method can avoid the data association problem in multi-sensor multi-object tracking. Moreover, it is more reliable and less computational than particle probability hypothesis density filter for multi-sensor multi-object tracking. We demonstrated the efficient of the approach by applications such as bearing and range tracking, and multiple speaker tracking.

Widely used in the evaluation of retrieval systems, the pooling method collects top ranked images from submitted retrieval systems resulting in possibly a very large pool of images. Inevitably, the pool may contain outliers. Human experts then manually annotate the relevance of them to create a ground truth for evaluation. Studies show that this annotation is time-consuming, tedious and inconsistent. To reduce human workload, this paper introduces an automatic method to detect outliers. Different from traditional detection methods using unsupervised techniques only, we utilize both supervised and unsupervised techniques sequentially as both positive and negative examples are (partially) available in this context. Specifically, support vector machines (SVMs) and fuzzy c-means clustering are used to predict data relevance and "outlierness". Performance improvements using our method after outlier removal have been validated on the medical image retrieval task in ImageCLEF 2004.

Most methods for multiple object tracking in video represent the state of multi-object in a high dimensional joint state space. This leads to high computational complexity. This paper presents a method using the probability hypothesis density (PHD) filter to estimate the state of multiple objects in video. The method operates on the single object state space instead of the joint state space. A PHD recursion for visual observations with color measurements is proposed. Our method can track varying number of objects.

In this paper, we present an efficient method for tracking multiple speakers in a reverberant environment. The proposed method is based on the cardinalized probability hypothesis density (CPHD) filter. Because the CPHD filter can handle a large amount of clutter measurements, our method has a high reliability when tracking multiple speakers. Simulation experiments are presented to demonstrate the performance of the proposed method.

Three-dimensional segmentation is reliable approach to achieve a proper estimation of tumor volume. Among all possible methods for this purpose, level set can be used as a powerful tool which implicitly extracts the tumor surface. In this paper, we introduce a threshold-based algorithm for 3D tumor segmentation using level set (TSL). This algorithm uses a global threshold to form the level set speed function which is updated iteratively throughout the level set growing process. An important feature of TSL is that no explicit knowledge about the tumor and non-tumor density functions is required. The proposed method can be implemented in the automatic and semi-automatic forms depending on the complexity of the tumor shape. TSL is examined on several clinical MRIs for both visual and quantitative evaluation. Experimental results demonstrate the effectiveness of our approach

Algorithms developed with a Gaussian noise assumption perform poorly in impulsive noise, such as that described by the symmetric alpha-stable (SalphaS) distribution. We investigate the performance of antipodal signaling and Viterbi decoding of convolutional codes in SalphaS noise. We demonstrate that the p-norm branch metric is robust in SalphaS noise.

Laser speckle imaging has increasingly become a viable technique for real-time medical imaging. However, the computational intricacies and the viewing experience involved limit its usefulness for real-time monitors such as those intended for neurosurgical applications. In this paper, we propose a new technique, tLASCA, which processes statistics primarily in the temporal direction using the laser speckle contrast analysis (LASCA) equation, proposed by Briers and Webster. This technique is thoroughly compared with the existing techniques for signal processing of laser speckle images, including, the spatial-based sLASCA and the temporal-based modified laser speckle imaging (mLSI) techniques. sLASCA is an improvement of the basic LASCA technique. In sLASCA, the derived contrasts are further averaged over a predetermined number of raw speckle images. mLSI, on the other hand, is the technique in which temporal statistics are processed using the equation described by Ohtsubo and Asakura. tLASCA preserves the original image resolution similar to mLSI. tLASCA outperforms sLASCA (window size M = 5) with faster convergence of K values (5.32 versus 20.56 s), shorter per-frame processing time (0.34 versus 2.51 s), and better subjective and objective quality evaluations of contrast images. tLASCA also outperforms mLSI with faster convergence of K values (5.32 s) compared to N values (10.44 s), shorter per-frame processing time (0.34 versus 0.91 s), smaller intensity fluctuations among frames (8%-10% versus 15%-35%), and better subjective and objective quality evaluations of contrast images. As laser speckle imaging becomes an important tool for real-time monitoring of blood flows and vascular perfusion, tLASCA is proven to be the technique of choice.

Occlusion or centric occlusion (CO) is the relationship between the upper and lower teeth when an individual bites his teeth together and is affected by the alignment and spacing of the upper and lower teeth. Research has indicated that CO has certain implications on temporo-mandibular joint (TMJ) dysfunction, and this usually leads to treatment such as complete denture reconstructions, TMJ therapy, orthodontics and facial pain treatment. In this paper, we attempt to develop a method of aligning the virtual dental cast models in CO so that one can study the CO intercuspal position (ICP). The method described here comprises two steps. First, we determine a good initial alignment using a plane detection strategy. Once the virtual dental models are in good initial alignment, we proceed to the second step of the alignment, which uses a collision detection strategy. The results obtained are within an error of 0.5mm which is acceptable for practical dental applications

Medical simulators are increasingly being used for surgical training. For interactive surgical simulation involving haptic rendering, the force at the needle tip has to be computed very fast. We are developing biomechanical models for bone needle insertion. The cortical bone can be regarded as a dense form of cancellous bone that can be modeled using a linear elastic material. The porosity of the bone determines the resistance felt as the user inserts the needle into the bone. The bar element method that represents each trabecular bone as a FE beam is most computationally efficient. With 1000 FE elements, the computed force feedback were close to the insertion force measured during experiments. However, the extended bar element method may be the more appropriate choice for taking into consideration the trabecular distribution and hence, inhomogeneous of bone. The simulation studies on bone-needle interaction also showed that a diamond bevel needle may penetrate the bone with less force

In this paper, we propose a filter bank architecture based on a frequency response masking design for variable bandwidth systems. The proposed method is shown to provide a greatly reduced complexity and delay for designs with sharp transition widths, compared to the conventional octave filter banks. Furthermore, the proposed architecture allows for greater flexibility in defining the bandwidths of each channel by combining channels with smaller bandwidths, and is also able to accommodate arbitrary decimation factors and is suitable for oversampled or critically-sampled systems. A 5-channel design example is provided in which we used a multiplierless design for the masking subfilters.

A novel similarity measure for edge-detection that is robust to varying luminance and contrast is presented. It incorporates a regularization term and employs directional FIR edge filters with hyperbolic tangent profiles to ensure improved noise performance and edge localization compared to classical methods.

The paper addresses the problem of deformable hand shape recognition in biometric systems without positioning aids. We separate and recognize multiple rigid fingers. An elliptical model is introduced to represent fingers and accelerate the matching of them. Technically, our method bridges the traditional handcrafted-feature methods and the shape-distance methods. We have tested it using our 108-person, 540-sample database with significantly increased positive recognition accuracy.

This paper proposes a content-based medical image retrieval (CBMIR) framework using dynamically optimized features from multiple regions of medical images. These regional features, including structural and statistical properties of color, texture and geometry, are extracted from multiple dominant regions segmented by applying Gaussian mixture modeling (GMM) and the expectation maximization (EM) algorithm to medical images. Over them, principal component analysis (PCA) is utilized to construct query templates and to reduce feature dimensions for representative feature optimization. Applying this method to the tasks of the medical imageCLEF 2004 we achieve better retrieval performance (MAP 0.4535) over the existing work on casImage of about 9000 images.

We propose a novel framework for content-based image retrieval with multiple parallel retrieval engines (MultiPRE) to achieve higher retrieval performance. Visual features, including both low-level features, such as color, texture and region features, and middle-level structure features, such as blob representation of objects are used to capture geometrical and statistical characteristics of images. Both clustering analysis and discrimination analysis are used as similarity measures in multiple retrieval engines, which are based on~principal component analysis (PCA) and support vector machines (SVM), respectively. Finally outputs of these engines are fused to determine ranking lists of retrieved images for given retrieval topics. The proposed framework has been evaluated based on the 26 image query topics over the CasImage database~with over 9000 medical images~used in ImageCLEF 2004, an international research effort for content-based image retrieval performance benchmark. Experiments show that the proposed framework achieved significantly better performance in terms of both the mean and the variance of average precision than the best run reported in ImageCLEF2004.

We describe a novel approach to obtain a good estimate of the complete 3D representation of a tooth given only the crown. The technique is based on the use of a statistical model derived from 3D images of teeth constructed by microtomography. The root is constructed by matching the respective crowns of the two models, defining the optimal registration and optimizing the statistical model parameters. This method allows us to generate the root efficiently due to the small number of parameters to adjust and requires little or no interaction from the user. The resultant 3D model is smooth and provides a good approximation of the root shape.

This paper presents the use of a PCA based approach to segment cells from RGB light microscope images. The proposed segmentation is accurate and robust under uneven illumination, lighting variation, and noise. Principal component analysis (PCA) is first applied to the RGB color bands of the image. The image corresponding to the principal component has significantly better contrast over the original image. A set of eigenfilters is then obtained by applying PCA to local neighborhoods of this image. A pair of filters from this set, corresponding to the second and third largest eigenvalues, resembles ramp edge filters with orientations that adapt to the image. These edge filters are used to obtain the edgemap of the image. We define a criterion that enables accurate detection of valid edges of cells while suppressing noise.

Contrast values estimated from temporal statistics is used for laser speckle contrast analysis (LASCA) of cortical blood flow and perfusion. Using temporal statistics, we are able to reliably estimate the blood flow and perfusion by processing only a selected number of pixels from the raw speckle images. For the number of frames (N > 6) for estimating the temporal contrast, the root mean square (RMS) of the inverse decorrelation time (/spl tau//sub c/) is found to be relatively constant with a variance of /spl plusmn/ 3.52%. For a given window size used for estimation of /spl tau//sub c/, it is observed that the RMS of 1/(/spl tau//sub c/ values exhibit a low variance of /spl plusmn/ 4.62% as compared to spatially derived contrasts (SDC) with/without averaging. For a given reduction in binning size, the image processing using the temporally derived contrasts (TDC) is found to be 1.41 times faster as compared to obtained using SDC.

The accurate segmentation of the teeth from the digitized representation of a dental study model is an important component in computer-based algorithms for orthodontic feature detection and measurement and in the simulation of orthodontic procedures such as tooth rearrangement. This paper presents an automated method for tooth segmentation from the three-dimensional (3-D) digitized image captured by a laser scanner. We avoid the complexity of directly processing 3-D mesh data by proposing the innovative idea of detecting features on two range images computed from the 3-D image. The dental arch is first obtained from the plan-view range image. Using the arch as the reference, a panoramic range image of the dental model can be computed. The interstices between the teeth are detected separately in the two range images, and results from both views are combined for a determination of interstice locations and orientations. Finally, the teeth are separated from the gums by delineating the gum margin. The algorithm was tested on 34 dental models representing a variety of malocclusions and was found to be robust and accurate.

In image matching no guarantee can be given that a solution exists, is unique, and is stable. This paper presents a method which utilises knowledge about the object present in two images to establish correspondence between the two images. In addition, the method uses the relative ordering of pixel intensities within a window, rather than the intensity values themselves. Experimental studies show that the proposed method is robust with respect to distortion, tolerates a small amount of random noise, and produces higher numbers of matched pixels than other similar matching methods.

The mobile agent paradigm is a promising technology for smart e-commerce of the next generation. However, security-related issues must be satisfactorily solved before they come into full commercial use. We propose a secure approach for mobile agent migration control based on passport and visa (P/V). P/V as up-to-date digital credentials not only serve the agent-host authentication purpose, but also can be used by hosts to trace the agent route and by communities to control the migration of mobile agents. The binding scheme further provides a way to protect these digital credentials and eliminate the misuse of them on illegal agents. We discuss the design issues in detail and evaluate the implementation of the proposed system.

Improvements in the compression performance can be achieved if the image transform is able to adapt to the shape of the objects present. We propose a new hybrid shape-adaptive orthogonal transform which is based on M-channel subband decomposition and shape-adaptive DCT techniques. The signal extension method for the region-based dyadic wavelet transform is applied to region-based image coding using M-channel filter banks. Simulation results show that the proposed method outperforms shape-adaptive DCT coding.

This paper describes a new method for image segmentation in which both pixel location and intensity similarity are taken into account. The proposed approach can be regarded as adaptive local thresholding since we focus on the analysis of local rather than global features. By applying the validity-guided fuzzy c-means algorithm, the spatial partitioning of an image into sub-regions in our method becomes context-oriented and fully automatic, unlike conventional local techniques. Experimental results indicate that the algorithm possesses robustness to uneven illumination, noise and presence of shadows.

A method for designing perfect-reconstruction M-channel integer-modulated filter banks (IMFBs) is proposed. Given the filter coefficient word length, many IMFBs satisfying the perfect reconstruction conditions can be obtained. Methods to select the best IMFB for the purpose of image compression are studied, and it is demonstrated that the IMFB obtained by maximizing the coding gain is the most suitable for image coding. Simulation results show that the PSNR of the integer-modulated FIR filter bank designed by maximizing the coding gain with the AR(1) model is close to that of the real-valued cosine-modulated filter bank of the same order.

Contrary to one of the key conclusions of the paper by Sikora and Makai (see ibid., vol.5, no.1, p.5962, 1995), the cross-correlation of the one-dimensional shape-adaptive DCT (1-D SA-DCT) coefficients located in two columns of different sizes does not necessarily take maximum values at the same indices. The correct way to measure the cross-correlation of the SA-DCT coefficients after the first stage of the transform is presented. The idea of implementing shape-adaptive DCT and some aspects of their work could nevertheless be useful.

A new class of morphological operators, the domain operator, is proposed for binary images. The operation is defined with a domain function based on the index function of fuzzy sets. With appropriate definition of the connection weights, the domain operator can implement conventional morphological operations such as dilation, erosion, opening, and closing, as well as other image processing functions such as noise reduction, edge detection, and clump splitting.