This paper proposes a new wavelet based image denoising method by using linear elementary parameterized denoising functions in the form of derivatives of Gaussian of a set of estimated wavelet coefficients. These coefficients are derived from an improved context modeling procedure in terms of mean square error estimation combining inter- and intra-sub band data. The denoising method results in a two-step denoising effort which outperforms the state-of-the-art non-redundant methods. This method is also extended to the over complete wavelet expansion by applying cycle spinning, which provides additional denoising performance and yields significantly better results than the orthogonal transform.

An eye-tracking integrated head-mounted display (ET-HMD) system can be used in many applications. Its complexity imposes great challenges on designing a compact, portable, and robust system. For an ET-HMD system, having accurate eye-gazed field-of-view (FOV) estimation and performing fast high-fidelity rendering only on the identified FOV is its crucial technique for augmented video/image/graphics display. In this paper, an energy-controlled iterative curve fitting method is proposed for accurate pupil detection and tracking. Based on a 3D eye model and the pupil tracking results, eye-gazed FOV can be identified precisely. The experimental results show that the average tracking error of using the proposed method is less than 0.5 degree for pupil rotation while the other conventional methods can only achieve 1.0 degree.

Three-dimensional (3D) data processing and modeling from range images plays an important role in many applications. Since overlap elimination of registered range images is a necessary step in 3D object modeling, many research efforts have been made. In this paper, a novel approach for eliminating overlaps of a registered data set is proposed. Firstly, the input data (registered range images) which contain overlaps is represented by a bd-tree data structure. Then, Moving Least Squares (MLS) and Spin Map (SM) together with the nearest neighbor searching are used to identify and eliminate the overlaps. The method manipulates the registered images directly without meshing them first, therefore, provides a straightforward way to remove redundant data, which makes it different from the conventional methods. The experimental results demonstrate the efficiency of the proposed algorithm.

This paper presents a new local edge-based level set model that does not use initial contours. Unlike traditional edge-based active contours that use gradient to detect edges, our model derives the neighborhood distribution and edge information with two different localized region-based operators: a Gaussian mixture model-based intensity distribution estimator and the Hueckel operator. We incorporate the operator outcomes into the recently proposed local binary fitting (LBF) model as local distribution fitting (LDF) model, which enables a model without the initial contour selection, i.e., the level set function can be initialized with a random constant instead of a distance map. Thus our model overcomes the initialization sensitivity problem of most active contours. In addition, with region-based edge detection, the proposed LDF model provides more accurate and robust segmentation. Experiments on both synthetic and real images show the improved performance of our proposed model over the LBF model.

This paper presents an efficient global optimization approach to the problem of constrained contour energy minimization for the object boundary extraction. In the method, with a given contour energy function, different target boundaries can be modeled as constrained global optimal solutions under different constraints expressed as a set of parameters characterizing the target contour interior structure. To search for the constrained global optimal solution, a fast and efficient global approach based on mean field annealing (MFA) is employed to avoid local minima. An illustrative example of three target boundaries in a synthetic image modeled as constrained global energy minimum contours with different constraint parameters is successfully located using the derived algorithm. A conventional variational based deformable contour method (Wang et al., 2002) with the same energy function and constraint fails to achieve the same task. Experimental evaluations and comparisons with other methods on ultrasound pig heart, MRI knee, and CT kidney images where gaps, blur contour segments having complex shape and inhomogeneous interiors have been conducted with most favorable results

This paper presents a robust and efficient skeleton-based graph matching method for object recognition and recovery applications. The novel feature is to unify both object recognition and recovery components into an image understanding system architecture, in which a complementary feedback structure can be incorporated to alleviate processing difficulties of each component alone. The idea is firstly to recognize the preliminary extracted object from a set of models using the new skeleton graph matching method, then to apply the a priori shape information of the identified model for accurate object recovery. The output of the system is the recognized and segmented object. The skeleton graph matching method is illustrated by recognizing a set of tool and animal silhouette examples with the presence of geometric transformations (translation, rotation, scaling, reflection), shape deformations and noise. Experiments of object recovery using MR knee images, have shown satisfactory results

We present a novel method to effectively segment the three dimensional MR brain images (volumes) with severe intensity nonuniformity. The segmentation problem was formulated using maximum a posterior probability and Markov random filed (MAP-MRF) framework. A novel spatial Gaussian mixture model (SGMM) is used to represent the intensity probability distribution of each of the three brain tissues (WM, GM and CSF), and MRF is used to compute the prior probability. This method consists of a learning process based on expectation maximization algorithm (EM) to estimate the parameters of SGMM, and a classification algorithm based on iterated conditional modes (ICM) to perform the segmentation of the sequential brain images using the parameters obtained from the learning process. The results on the simulated and twenty in vivo MR brain volumes demonstrate the efficiency of this method. We also present the comparison results with other published methods.

A divide and conquer deformable contour method is presented with an initial inside closed contour being divided into arbitrary segments, and these segments are allowed to deform separately preserving the segments' connectivity. A maximum area threshold, A/sub max/, is used to stop these outward contour segments' marching. Clear and blur contour points are then identified to partition the whole contour into clear and blur segments. A bi-directional searching method is then recursively applied to each blur segment including a search for contour-within-contour segment to reach a final close contour. Further improvements are provided by a model based searching algorithm. It is a two-step process with step 1 being a linked contour model matching operation where landmarks are extracted, and step 2 being a posteriori probability model matching and correction operation where large error segments are fine tuned to obtain the final results. The experiments include ultrasound images of pig heart, MRI brain images, MRI knee images having complex shapes with or without gaps, and inhomogeneous interior and contour region brightness distributions. These experiments have shown that the method has the capability of moving a contour into the neighboring region of the desired boundary by overcoming inhomogeneous interior, and by adapting each contour segment searching operation to different local difficulties, through a contour partition and repartition scheme in searching for a final solution.

A robust skeleton-based shape matching method for model-based shape recovery applications is presented. The object model consists of both a skeleton model and a contour segments model, which are used in tandem and in a complementary manner. Initially, the skeleton of the contour, provided by a deformable contour method (DCM), is matched against a set of object skeleton models to select a candidate model and determine the corresponding landmarks on the contours. Segments obtained from these landmarks are then matched against the detected model segments for errors. For any large segment mismatch error, a fine-tuning process, which is formulated as a maximization of a posteriori probability, given the contour segments model and image features, is performed for the final result. The skeleton matching algorithm is illustrated by using a set of animal profile examples. Experimental results of shape recovery from practical applications, such as an MR knee image, are very encouraging.

The improvements in structured lighting based 3D optical camera measurement systems have made these non-contact inspection systems more applicable. In this paper, a basic calibration problem, caused by systematical errors particular to structured lighting based 3D optical camera systems, is addressed. The analysis of these errors is discussed and two model-based approaches, line model (LM) and area model (AM), are proposed to model the pattern of systematical error. The main step of either approach is to build a lookup table based on collected planar data at predetermined working distances and orientation angles. The lookup table is then used to correct the systematical error existing in the inspected data. The experimental results show that the proposed approaches effectively improve inspection accuracy.

In this paper, a deformable contour method is formulated and derived in a constrained optimization framework. The constraint is reflected in the resulting contour evolution equation. This constraint element can be used to characterize any interior property and/or structure including the added "balloon force" used in geodesic snake as a special case, which has the potential of having expanded applications to far more complex interior. An implementation of the method is applied to a MRI brain image and the results are favorable.

A two-step model based approach to a contour extraction problem is developed to provide a solution to more challenging contour extraction problems of biomedical images. A biomedical contour image is initially processed by a deformable contour method to obtain a first order approximation of the contour. The two-step model includes a linked contour model and a posteriori probability model. Initially, the output contour from the deformable contour method is matched against the linked contour model for both model detection and corresponding landmark contour points identification. Segments obtained from these landmarks are matched for errors. Larger error are then passed on to a regionalized a posteriori probability model for further fine tuning to obtain a final result. Experiments on both MR brain images are most encouraging.

A divide and conquer strategy in the deformable contour method is presented. An initial inside closed contour is divided into segments, and these segments are allowed to deform separately preserving segments' connectivity. A deformable contour algorithm is adapted to each contour segment movement. A maximum area threshold, A/sub max/, is used to stop these outward contour segment marchings. Clear and blur contour points are then identified, and the whole contour is repartitioned into clear, blur, and gap segments. A bi-directional searching method is then recursively applied to each, blur, or gap segment until a final contour is sought. At this point, a search for contour within contour segment is undertaken so that the inner most contour can be searched. At all times, a global snake type performance index is used to find each local contour segment. Experiments have shown that the method has the capability of moving a contour into the neighboring region of the solution contour by overcoming all the above inhomogeneous interior, and of adapting each contour segment searching operation to different local difficulties, through a contour partition and repartition scheme in searching for a final solution. These experiments include ultrasound images of pig heart, MRI brain images, and MRI knee images having complex shapes and/or with gaps, inhomogeneous interior and contour region brightness distributions.

Several adaptive order statistic filters (OSF) are developed and compared for channel characterization and noise suppression in images. Emphasis has been put on the situation when a noise-free reference image is not available but instead we can have a sequence of two noisy output images of the same input image through the communication channel twice. One of the noisy output images is used as the reference in the OSF. It is shown theoretically that if noises are not correlated, the expected values of the derived filter coefficients will be equal to those coefficients derived using a noise-free reference. Experiments using noisy reference image yield comparable results to those methods using a noise-free reference image and also better results than those of median, Gaussian, averaging and Wiener filters.

Wastewater treatment system design can be identified as a planning problem. The wastewater consists of several chemicals (compounds) that need to be removed during the treatment process. A treatability database containing the treatability of various compounds through different types of treatment processes has been developed. In general, two or more compounds appear together, and one may need to combine and arrange two or more treatment processes to meet the treatment objectives. The wastewater treatment system has been designed in two phases. In the first phase, an inductive learning algorithm is used to extract the knowledge rules from the database. These rules are compiled to identify the effect of an individual treatment process on several compounds at different concentrations. The second phase involves selecting, combining, and arranging the unit treatment processes that will meet all treatment objectives. This phase has been formulated as an optimization problem. A Hopfield network is used to obtain the necessary sequence of treatment processes to achieve the desired level of effluent concentration from the given influent concentrations

A planning model described in terms of its goal analysis and hierarchical operator representation is presented. With this system, successful plans have been made for nonlinear problems that are described as a conjunction of subgoals. The system uses heuristic rules to analyze the problem, and thus achieves an ordered sequence of subgoals and constraints that can be achieved successively without interfering with each other. The operators are designed in a goal-oriented fashion and are stored in operator hierarchies. During the plan generation phase, each subgoal is mapped into a goal operator, which is further refined to meet details and specific conditions of the problem. As the generation phase follows the analysis, conflicts among subgoals are eliminated implicitly.>

Our goal is to develop a vision system using range image for the industrial application. In the system, simple basic surface primitives are considered and are extracted from the range image. From the robust geometric features contained in the primitives, rotation-invariant characteristics are then extracted to form smallest distinctive feature sets (SDFS). The set of SDFSs and the relationships between elements in SDFS and the kernel are stored to form the object model. The recognition process involves first extracting a set of SDFSs from the image and matching the extracted SDFSs with the stored SDFSs. If the kernels determined for SDFSs matched are consistent, object can be detected and its location and orientation in the viewer-centered coordinate system (VCC) can be determined.

This paper describes an approach to generate task level plans for mechanical assembly using robots. The system generates assembly plans by inferring from the knowledge base of workpiece structures, assembly rules, and robot operations. The novel feature of the system is its goal (workpiece) structure analysis. The goal structure analysis is capable of producing a correctly ordered sequence of subgoals such that the interactions among the subgoals will not interfere with achieving the goal. A plan generator then follows the analysis information to select the appropriate robot operations to formulate the final plan without much searching and reasoning. We believe that this goal-analyzing approach is much closer to human-like planning model than other interaction-resolving planning systems.

The original algorithm of Hough transform was for detecting and finding straight lines in two dimensional images. Since then, it has been extended to detect and locate special planar curves such as circles, parabolas, etc. Recently, generalized Hough transform has been applied for recognizing and locating three dimensional objects using range data. Our work is based on the same principle by decomposing the overall parameter set of nine elements representing a quadric into three decoupled parameter subsets: scalar parameter, translational parameter, and orientational parameters. The use of Hough transform on each decoupled subset in an independent but sequential manner rather than the whole enables us to reduce the computational cost and storage requirements by orders of magnitude as compared to all previous published applications.

The three-dimensional reconstruction of the radioisotope source distribution of coordinated radioisotope scintigrams is complicated by the point spreading function created by the imaging system (conical imaging) and the attenuation of gamma rays through the body. A new reconstruction method is proposed by converting the complicated existing reconstruction problem into the classical problem of weighted projection in the Fourier complex space upon which a modified additive algebraic reconstruction technique (ART) can be employed. Mathematical derivation, reconstruction algorithms, and illustrative computer results are included.

Projection transform technique is applied to digital image transmission. Transform samples on various central sections of a Fourier space are attained with a fast Fourier transform algorithm. Quantization, coding, and transmission follow in order. The complete Fourier space information is then, upon reception of signals, estimated from the transform samples at these various central sections. The reconstructed picture is produced by the two-dimensional inverse Fourier transform.

This paper gives a unified approach to designing a data analyzer that performs cluster-seeking, feature selection, and categorizer design under a weighted least-square performance criterion. The cost of misrecognitions is preserved throughout the process, It can be used as a fast procedure to evaluate the discriminatory capability of sensors and/or preprocessors.

A feature-selection procedure is proposed for the class of distribution-free pattern classifiers [1], [2]. The selection procedure can be readily carried out on fixed (large) training samples using matrix inversion. If direct matrix inversion is to be avoided, the approximation method [4] or the stochastic-approximation procedure [2] can be applied to the training samples. The proposed procedure, aside from furnishing a statistical interpretation, has a mapping interpretation. It has the unique property of designing a pattern classifier under a single-performance criterion instead of the conventional division of receptor and categorizer. It enables the system to come closest to the minimum-risk ideal classifier. In particular, for two-class problems having normal distributions with equal covariance matrices, equal costs for misrecognition, and equal a priori probabilities, the proposed procedure yields the optimum Bayes procedure without the knowledge of the class distributions. Furthermore, the proposed feature-selection procedure is the same as that of the divergence computation. Experimental results are presented. They are considered satisfactory.

Abstract—In this paper a least-square approach to multiclass pattern classification is undertaken. The generalized inverse computation is used to furnish a quick solution to the problem of fixed training samples. The use of recursive on-line computation is also recommended. Experimental results are presented to illustrate the approach. Both deterministic and statistical interpretations have been given to the approach. The pattern classifier proposed by Chaplin and Levadi [1] and the adaptive pattern classifier proposed by Patterson and Womack [2] are special cases of this approach.

Abstract—A more general adaptive procedure for determining linear or piecewise linear discriminant functions for multiclass pattern classification is proposed. The adaptive procedure is a many-pattern or group-pattern adaptation. The training sequence consists of groups of vectors in matrix form instead of single vectors. The convergence proof shows that this procedure terminates in a finite number of adaptions if the solution exists. A necessary and suffcient condition is developed for testing the linear separability of each subset of (d + 1) samples. Furthermore, the proposed procedure can be implemented with the addition of only a little complexity to existing systems. Computer simulations indecate satisfactory results.