Fadi Chakik

Efficient object detection and tracking in video sequences.

One of the most important problems in computer vision is the computation of the two-dimensional projective transformation (homography) that maps features of planar objects in different images and videos. This computation is required by many applications such as image mosaicking, image registration, and augmented reality. The real-time performance imposes constraints on the methods used. In this paper, we address the real-time detection and tracking of planar objects in a video sequence where the object of interest is given by a reference image template. Most existing approaches for homography estimation are based on two steps: feature extraction (first step) followed by a combinatorial optimization method (second step) to match features between the reference template and the scene frame. This paper has two main contributions. First, we detect both planar and nonplanar objects via efficient object feature classification in the input images, which is applied prior to performing the matching step. Second, for the tracking part (planar objects), we propose a fast method for the computation of the homography that is based on the transferred object features and their associated local raw brightness. The advantage of the proposed schemes is a fast matching as well as fast and robust object registration that is given by either a homography or three-dimensional pose.

Efficient Object Detection and Matching Using Feature Classification

This paper presents a new approach for efficient object detection and matching in images and videos. We propose a stage based on a classification scheme that classifies the extracted features in new images into object features and non-object features. This binary classification scheme has turned out to be an efficient tool that can be used for object detection and matching. By means of this classification not only the matching process becomes more robust and faster but also the robust object registration becomes fast. We provide quantitative evaluations showing the advantages of using the classification stage for object matching and registration. Our approach could lend itself nicely to real-time object tracking and detection.

Fraud Detection in Mobile Telecommunication

—This paper presents a predictive model to handle customer insolvency in advance for large mobile telecommunication companies for the purpose of minimizing their losses. However, another goal is of the highest interest for large mobile telecommunication companies is based on maintaining an overall satisfaction of the customers which may have important consequences on the quality and on the consume return of the operations. In this paper, the customer insolvency is defined to be a classification problem since our main purpose is to categorize the customer in one of the two classes: potentially insolvent or potentially solvent. Therefore, a model with precise business prediction using the knowledge discovery and Data Mining techniques on an enormous heterogeneous and noisy data is proposed. Moreover, a fuzzy approach to evaluate and analyze the customer behavior leading to segment them into groups that provide better understanding of customers is developed. These groups with many other significant variables feed into a classification algorithm to classify the customers. A real case study is considered here, followed by analysis and comparison of the results for the reason to select the best classification model that maximizes the accuracy for insolvent customers and minimizes the error rate in the misclassification of solvent customers.

Image compression based on fuzzy segmentation and anisotropic diffusion

In this paper we present a hybrid model for image compression based on fuzzy segmentation and Partial Differential Equations. The main motivation behind our approach is to produce immediate access to objects/features of interest in a high quality decoded image which could be useful on smart devices, for analysis purpose, as well as for multimedia content-based description standards. The image is approximated as a set of uniform regions: The technique will assign well-defined members to homogenous regions in order to achieve image segmentation. The fuzzy c-means (FcM) is a guide to cluster image data. A second stage coding is applied using entropy coding to remove the whole image entropy redundancy. In the decoding phase, we suggest the application of a nonlinear anisotropic diffusion to enhance the quality of the coded image.

Complexity Reduction and Quality Enhancement in Image Coding

In this paper we propose an image coding approach based on Alternative Fuzzy c-Means. Our main objective is to provide an immediate access to targeted features of interest in a high quality decoded image. This technique is useful for intelligent devices, as well as for multimedia content-based description standards. The use of AFcM reduces the coding time in comparison to the traditional clustering algorithm FcM. A second stage coding is applied using entropy coding to remove the whole image entropy redundancy. In the decoding phase, we suggest the application of a nonlinear anisotropic diffusion, based on Perona-Malik equation, to enhance the quality of the coded image. Qualitative evaluation confirms the validity of the proposed approach.

Linear vs. non-linear dimensionality reduction techniques in predicting class-II MHC peptide binding

A key step in the development of an adaptive immune response to vaccines is the binding of peptides to molecules of the Major Histocompatibility Complex (MHC) for presentation to T lymphocytes, which are thereby activated. Several algorithms have been proposed for such binding predictions, but are limited to a small number of MHC molecules and present imperfect prediction power. We are undertaking an exploration of the power gained by taking advantage of a natural representation of the protein sequence amino acid in terms of their composition, structural and a series of associated physicochemical properties to form a representative descriptor vectors. We are proposing to use dimensionality reduction techniques to preprocess the descriptor vectors before feeding them into well known statistical classifiers for binding prediction. In all cases, coupling dimensionality reduction techniques with the physicochemical properties leads to substantially higher values for our evaluation criteria (Area Under ROC Curve) which means that misclassification errors is reaching lower rates.

Real-time object detection and tracking in video sequences

One of the most important problems in Computer Vision is the computation of the 2D projective transformation (homography) that maps features of planar objects in different images and videos. This computation is required by many applications such as image mosaicking, image registration, and augmented reality. The real-time performance imposes constraints on the methods used. In this paper, we address the real-time detection and tracking of planar objects in a video sequence where the object of interest is given by a reference image template. Most existing approaches for homography estimation are based on two steps: feature extraction (first step) followed by a combinatorial optimization method (second step) to match features between the reference template and the scene frame. This paper has two main contributions. First, for the detection part, we propose a feature point classification which is applied prior to performing the matching step in the process of homography calculation. Second, for the tracking part, we propose a fast method for the computation of the homography that is based on the transferred object features and their associated local rawbrightness. The advantage of this proposed scheme is a fast and accurate estimation of the homography.