Kamel Hamrouni

3D SEGMENTATION OF MRI BRAIN USING LEVEL SET AND UNSUPERVISED CLASSIFICATION

The precise volumetric brain segmentation of normal and pathological MR images is a challenging problem in understanding brain anatomy and functions. The requirement to automate this process is another challenge because manual segmentation is extremely laborious. However, it has been proven to be problematic, both due to the high complexity of anatomical structures as well as their large variability, and due to difficulties caused by the presence of noise and artefacts. This paper describes a fully automatic and accurate method for segmenting normal brain tissues and a semi-automatic procedure for delineating meningiomas tissues. The approach performs the segmentation using a succession operations involving a registration step from known data, a classification step and a segmentation step based on level-set. The role of the registration and the classification is to initialize accurately the active model and to control its evolution. A hybrid region-boundary model using level set technique is also proposed. Our formalism is robust to noise and intensity inhomogeneities, since it exploits the advantage of the combination. It is evaluated on both simulated and real data, and compared with existing segmentation techniques. Qualitative and quantitative results demonstrated its performance and robustness.

Copy-move image forgery detection based on SIFT descriptors and SVD-matching

In recent years, more data are created in digital form allowing easy control over storage and manipulation due to the technology progress. Unfortunately, this progress may have dragged along a lot of risks, especially the ones related to the security of digital files. In particular, digital forgery becomes a worry for many organizations because it has become easier to create fake images without leaving any obvious perceptual traces of tampering. A specific form of image forgery operation called “copy-move” is considered one of the most difficult problems in the case of forgery detection. For this case, a part of the image is copied and pasted on another location of the same image to conceal undesirable objects in the scene. In this paper, we propose a method that automatically detects duplicated regions in the same image. Duplicated detection is performed by identifying the local characteristics of the images (points of interest) using the Scale Invariant Feature Transform (SIFT) method and by matching between identical features using the Singular Value Decomposition (SVD) method. Obtained results show that our proposed hybrid method is robust to geometrical transformations and is able to detect with high performance duplicated regions.

Non-parametric depth calibration of a TOF camera

Time-of-Flight (TOF) cameras measure, in real-time, the distance between the camera and objects in the scene. This opens new perspectives in different application fields: 3D reconstruction, Augmented Reality, video-surveillance, etc. However, like any sensor, TOF cameras have limitations related to their technology. One of them is distance distortion. In this paper, we present a new depth calibration method (estimation of distance distortion) for TOF cameras. Our approach has several advantages. First, it is based on a non-parametric model, contrary to most of the other methods. Second, it models under the same formalism the distortion variation according to the distance and the pixel position in the image. This improves calibration accuracy even at the image boundaries which are typically more distorted than the image center. A comparison with two state of the art parametric methods is presented.

Improving Iris Recognition Performance Using Quality Measures

Biometric methods, which identify people based on physical or behavioural characteristics, are of interest because people cannot forget or lose their physical characteristics in the way that they can lose passwords or identity cards. Among these biometric methods, iris is currently considered as one of the most reliable biometrics because of its unique texture‘s random variation. Moreover, iris is proved to be well protected from the external environment behind the cornea, relatively easy to acquire and stable all over the person’s life. For all of these reasons, iris patterns become interesting as an alternative approach to reliable visual recognition of persons. This recognition system involves four main modules: iris acquisition, iris segmentation and normalization, feature extraction and encoding and finally matching. However, we noticed that almost all the iris recognition systems proceed without controlling the iris image’s quality. Naturally, poor image’s quality degrades significantly the performance of the recognition system. Thus, an extra module, measuring the quality of the input iris, must be added to ensure that only “good iris” will be treated by the system. The proposed module will be able to detect and discard the faulty images obtained in the segmentation process or which not have enough information to identify person. In literature, most of evaluation quality methods have developed indices to quantify occlusion, focus, contrast, illumination and angular deformation. These measurements are sensitive to segmentation errors. Only few methods have interested on the evaluation of iris segmentation. This chapter aims to present, firstly a novel iris recognition method based on multi-channel Gabor filtering and Uniform Local Binary Patterns (ULBP), then to define a quality evaluation method which integrates additional module to the typical recognition system. Proposed method is tested on Casia v3 iris database. Our experiments illustrate the effectiveness and robustness of ULBP to extract rich local and global information of iris texture when combined with simultaneously multi-blocks and multi-channel method. Also, obtained results show an improvement of iris recognition system by incorporating proposed quality measures in the typical system. This chapter is organized as follows: Section 2 describe in details the proposed iris recognition system. The further represents the quality evaluation method. In section 4, we expose experiments, results and comparison. Finally, the conclusion is given in section 5.

Using multiple steerable filters and Bayesian regularization for facial expression recognition

Abstract Facial expression recognition has recently become a challenging research area. Its applications include human–computer interfaces, human emotion analysis, and medical care and cure. In this paper, we present a new challenging method to recognize seven universal emotional expressions, which are happiness, neutral, angry, disgust, sadness, fear and surprise. In our approach, we identify the user׳s facial expressions from the input images, using statistical features extracted from the steerable pyramid decomposition, and classified with a Bayesian regularization neural network. The evaluation of the proposed approach in terms of recognition accuracy is achieved using two universal databases, the Japanese Female Facial Expression database and the Cohn–Kanade facial expression database. The overall accuracy rate reaches 93.33% for the first database and is about 98.13% for the second one. These results show the effectiveness of the steerable proposed algorithm.

A Novel Content Image Retrieval Method Based on Contourlet

Content-based image retrieval is an active and fast advancing research area since the 1990s as a result of advances in the Internet and new digital image sensor technologies. However, many challenging research problems continue to attract researchers from multiple disciplines. Content-based image retrieval uses the visual contents of an image as features to represent and index the image to be searched from large scale image databases. The quality of the selected features relies mainly on the degree of the invariance property that is ensured under acceptable manipulations. This paper proposes an efficient method for compactly representing color and texture features and combining them for image retrieval. The performance of retrieval based on these compact descriptors obtained by the proposed techniques is analyzed and tested on wang database images yielding satisfactory accuracy rates. A comparative study demonstrated that the developed feature extraction scheme outperformed the other schemes being compared with.

Automatic MRI Brain Segmentation with Combined Atlas-Based Classification and Level-Set Approach

The task of manual brain segmentation from magnetic resonance imaging (MRI) is generally time-consuming and difficult. In a previous paper [1], we described a method for segmenting MR which is based on EM algorithm and a deformable level-set model. However, this method was not fully automatic. In this paper, we present an automated approach guided by digital anatomical atlas, which is an additional source of prior information. Our fully automatic method segments white matter, grey matter and cerebrospinal-fluid. The paper describes the main stages of the method, and presents preliminary results which are very encouraging for clinical practice.

Iris Recognition using Steerable Pyramids

This work presents a new iris recognition method based on steerable pyramid transform. This method consists of four steps: localization, normalization, features extraction and matching. After locating the iris boundaries by Hough Transform, normalization is operated by unwrapping the circular ring and isolating the noisy regions. Steerable pyramid filters are then used to capture orientation details from the iris texture. The features are extracted on each filtered sub-image to form a fixed length feature vector which will be compared to other vectors in the matching step. This technique has been tested on infrared light iris images. It has been compared, in both identification and verification modes, to known methods.

Image and video denoising by combining unsupervised bounded generalized gaussian mixture modeling and spatial information

In recent years, a great deal of effort has been expended on developing robust solutions for images quality degradation caused mainly by noise. In this paper, we explore this area of research and we propose a new unsupervised algorithm for both image and video denoising. Our solution is based on a flexible statistical mixture model driven by a finite mixtures of bounded generalized Gaussian distributions (BGGMD) which offers more flexibility in data modeling than the well known classical gaussian distributions which fail to fit the shape of heavy-tailed data produced by the presence of noise or outliers. The proposed framework takes into account also spatial information between neighboring pixels to be more robust and flexible, and able to provide smooth and accurate denoising results. For model’s parameters estimation, we investigate the unsupervised expectation-maximization (EM) algorithm. In order to evaluate the performance of the proposed model, we conducted a series of extensive experiments. Obtained results are more encouraging than those obtained using similar approaches. These results show the robustness and flexibility of the proposed method to adapt different shapes of observed data and bounded support data in the case of noisy images and videos.

Noise modelling in time-of-flight sensors with application to depth noise removal and uncertainty estimation in three-dimensional measurement

Time-of-flight (TOF) sensors provide real-time depth information at high frame-rates. One issue with TOF sensors is the usual high level of noise (i.e. the depth measures repeatability within a static setting). However, until now, TOF sensors’ noise has not been well studied. The authors show that the commonly agreed hypothesis that noise depends only on the amplitude information is not valid in practice. They empirically establish that the noise follows a signal-dependent Gaussian distribution and varies according to pixel position, depth and integration time. They thus consider all these factors to model noise in two new noise models. Both models are evaluated, compared and used in the two following applications: depth noise removal by depth filtering and uncertainty (repeatability) estimation in three-dimensional measurement.