Moumen T. El-Melegy

On Cluster Validity Indexes in Fuzzy and Hard Clustering Algorithms for Image Segmentation

This paper addresses the issue of assessing the quality of the clusters found by fuzzy and hard clustering algorithms. In particular, it seeks an answer to the question on how well cluster validity indexes can automatically determine the appropriate number of clusters that represent the data. The paper surveys several key existing solutions for cluster validity in the domain of image segmentation. In addition, it suggests two new indexes. The first one is based on Akaikes information criterion (AIC). While AIC was devoted to other domains such as statistical estimation of model fitting, it is implemented here for the first time as a validation index. The second index is developed from the well-established idea of cross-validation. The existing and new indexes are evaluated and compared on several synthetic images corrupted with noise of varying levels and volumetric MR data.

Robust Training of Artificial Feedforward Neural Networks

Artificial feedforward neural networks have received researchers’ great interest due to its ability to approximate functions without having a prior knowledge about the true underlying function. The most popular algorithm for training these networks is the backpropagation algorithm that is based on the minimization of the mean square error cost function. However this algorithm is not robust in the presence of outliers that may pollute the training data. In this chapter we present several methods to robustify neural network training algorithms. First, employing a family of robust statistics estimators, commonly known as M-estimators, in the backpropagation algorithm is reviewed and evaluated for the task of function approximation and dynamical model identification. As theseM-estimators sometimes do not have sufficient insensitivity to data outliers, the chapter next resorts to the statistically more robust estimator of the least median of squares, and develops a stochastic algorithm to minimize a related cost function. The reported experimental results have indeed shown the improved robustness of the new algorithm, especially compared to the standard backpropagation algorithm, on datasets with varying degrees of outlying data.

Development of a Micro-Gripper Using Piezoelectric Bimorphs

Piezoelectric bimorphs have been used as a micro-gripper in many applications, but the system might be complex and the response performance might not have been fully characterized. In this study the dynamic characteristics of bending piezoelectric bimorphs actuators were theoretically and experimentally investigated for micro-gripping applications in terms of deflection along the length, transient response, and frequency response with varying driving voltages and driving signals. In addition, the implementation of a parallel micro-gripper using bending piezoelectric bimorphs was presented. Both fingers were actuated separately to perform mini object handling. The bending piezoelectric bimorphs were fixed as cantilevers and individually driven using a high voltage amplifier and the bimorph deflection was measured using a non contact proximity sensor attached at the tip of one finger. The micro-gripper could perform precise micro-manipulation tasks and could handle objects down to 50 μm in size. This eliminates the need for external actuator extension of the microgripper as the grasping action was achieved directly with the piezoelectric bimorph, thus minimizing the weight and the complexity of the micro-gripper.

Tumor segmentation in brain MRI using a fuzzy approach with class center priors

This paper proposes a new fuzzy approach for the automatic segmentation of normal and pathological brain magnetic resonance imaging (MRI) volumetric datasets. The proposed approach reformulates the popular fuzzy c-means (FCM) algorithm to take into account any available information about the class center. The uncertainty in this information is also modeled. This information serves to regularize the clusters produced by the FCM algorithm thus boosting its performance under noisy and unexpected data acquisition conditions. In addition, it also speeds up the convergence process of the algorithm. Experiments using simulated and real, both normal and pathological, MRI volumes of the human brain show that the proposed approach has considerable better segmentation accuracy, robustness against noise, and faster response compared with several well-known fuzzy and non-fuzzy techniques reported in the literature.

SHREC'13 track: retrieval on textured 3D models

This contribution reports the results of the SHREC 2013 track: Retrieval on Textured 3D Models, whose goal is to evaluate the performance of retrieval algorithms when models vary either by geometric shape or texture, or both. The collection to search in is made of 240 textured mesh models, divided into 10 classes. Each model has been used in turn as a query against the remaining part of the database. For a given query, the goal was to retrieve the most similar objects. The track saw six participants and the submission of eleven runs.

Random Sampler M-Estimator Algorithm With Sequential Probability Ratio Test for Robust Function Approximation Via Feed-Forward Neural Networks

This paper addresses the problem of fitting a functional model to data corrupted with outliers using a multilayered feed-forward neural network. Although it is of high importance in practical applications, this problem has not received careful attention from the neural network research community. One recent approach to solving this problem is to use a neural network training algorithm based on the random sample consensus (RANSAC) framework. This paper proposes a new algorithm that offers two enhancements over the original RANSAC algorithm. The first one improves the algorithm accuracy and robustness by employing an M-estimator cost function to decide on the best estimated model from the randomly selected samples. The other one improves the time performance of the algorithm by utilizing a statistical pretest based on Walds sequential probability ratio test. The proposed algorithm is successfully evaluated on synthetic and real data, contaminated with varying degrees of outliers, and compared with existing neural network training algorithms.

Neural Networks in Multiple Classifier Systems for Remote-Sensing Image Classification

Summary In recent years there has been a growing interest in the development of super-vised classification techniques with higher classification reliability of satelliteimages. The superiority of one technique over the others cannot be claimed.Many experimental results showed that the classification accuracy dependsmore on the particular application than on the technique chosen to performthe task. Moreover in many applications it is very difficult to design a clas-sification system that exhibits the required accuracy for the final classifica-tion product. Therefore a new technique is emerging that considers multipleclassifier systems (MCSs) instead of a single classification technique. Neuralnetworks can participate effectively in a MCS in several ways. This chapterfocuses on the role of neural networks in MCSs, which can be either: (1) anindividual classifier among the classifiers ensemble, (2) the fusion center thatintegrates the decisions of individual classifiers, (3) the selector that pickssome classifiers’ decisions and ignores the others, (4) or the selector and fuserat the same time.The chapter surveys key neural and non-neural methods in this area anddevelops some new ones. It also considers several architectures of neural net-works including traditional feed-forward neural networks, probabilistic neuralnetworks and radial-basis nets. All methods are evaluated on two real remote-sensing datasets, one of which is the standard benchmark Satimage dataset.

RANSAC algorithm with sequential probability ratio test for robust training of feed-forward neural networks

This paper addresses the problem of fitting a functional model to data corrupted with outliers using a multilayered feed-forward neural network (MFNN). Almost all previous efforts to solve this problem have focused on using a training algorithm that minimizes an M-estimator based error criterion. However the robustness gained from M-estimators is still low. Using a training algorithm based on the RANdom SAmple Consensus (RANSAC) framework improves significantly the robustness of the algorithm. However the algorithm typically requires prolonged period of time before a final solution is reached. In this paper, we propose a new strategy to improve the time performance of the RANSAC algorithm for training MFNNs. A statistical pre-test based on Walds sequential probability ratio test (SPRT) is performed on each randomly generated sample to decide whether it deserves to be used for model estimation. The proposed algorithm is evaluated on synthetic data, contaminated with varying degrees of outliers, and have demonstrated faster performance compared to the original RANSAC algorithm with no significant sacrifice of the robustness.

Heat Kernels for Non-Rigid Shape Retrieval: Sparse Representation and Efficient Classification

One of the major goals of computer vision and machine intelligence is the development of flexible and efficient methods for shape representation. This paper presents an approach for shape retrieval based on sparse representation of scale-invariant heat kernel. We use the Laplace-Beltrami eigen functions to detect a small number of critical points on the shape surface. Then a shape descriptor is formed based on the heat kernels at the detected critical points for different scales, combined with the normalized eigen values of the Lap lace-Beltrami operator. Sparse representation is used to reduce the dimensionality of the calculated descriptor. The proposed descriptor is used for classification via the collaborative representation-based classification with regularized least square algorithm. We compare our approach to two well-known approaches on two different data sets: the nonrigid world data set and the SHREC 2011. The results have indeed confirmed the improved performance of the proposed approach, yet reducing the time and space complicity of the shape retrieval problem.

Statistically Robust Approach to Lens Distortion Calibration with Model Selection

This paper addresses the problem of calibrating camera lens distortion, which can be significant in medium to wide angle lenses. While almost all existing nonmetric distortion calibration methods need user involvement in one form or another,we present an approach to distortion calibration based on the robust the-least-median-of-squares (LMedS) estimator. Our approach is thus able to proceed in a ful ly-automatic manner while being less sensitive to erroneous input data such as image curves that are mistakenly considered as projections of 3D linear segments. Our approach uniquely uses fast, closed-form solutions to the distortion coefficients, which serve as an initial point for a non-linear optimization algorithm to straighten imaged lines. Moreover we propose a method for distortion model selection based on geometrical inference.Successful experiments to evaluate the performance of this approach on synthetic and real data are reported.