Eslam A. Mostafa

Face recognition in low resolution thermal images

This paper proposes an accurate, rotation invariant, and fast approach for detection of facial features from thermal images. The proposed approach combines both appearance and geometric information to detect the facial features. A texture based detector is performed using Haar features and AdaBoost algorithm. Then the relation between these facial features is modeled using a complex Gaussian distribution, which is invariant to rotation. Experiments show that our proposed approach outperforms existing algorithms for facial features detection in thermal images. The proposed approachs performance is illustrated in a face recognition framework, which is based on extracting a local signature around facial features. Also, the paper presents a comparative study for different signature techniques with different facial image resolutions. The results of this comparative study suggest the minimum facial image resolution in thermal images, which can be used in face recognition. The study also gives a guideline for choosing a good signature, which leads to the best recognition rate.

A novel probabilistic simultaneous segmentation and registration using level set

We propose a new shape-based segmentation approach using the statistical shape prior and level sets method. The segmentation depends on the image information and shape prior. Training shapes are grouped to form a probabilistic model. The shape model is embedded into the image domain taking in consideration the evolution of a contour represented by a level set function. The evolution of the front gathers information from the image intensities and shape prior. The segmentation approach is applied in segmenting the vertebral bodies in CT images. Our results shows that the technique is accurate and robust compared with the other alternative in the literature.

Pose invariant approach for face recognition at distance

We propose an automatic pose invariant approach for Face Recognition At a Distance (FRAD). Since face alignment is a crucial step in face recognition systems, we propose a novel facial features extraction model, which guides extended ASM to accurately align the face. Our main concern is to recognize human faces under uncontrolled environment at far distances accurately and fast. To achieve this goal, we perform an offline stage where 3D faces are reconstructed from stereo pair images. These 3D shapes are used to synthesize virtual 2D views in novel poses. To obtain good synthesized images from the 3D shape, we propose an accurate 3D reconstruction framework, which carefully handles illumination variance, occlusion, and the disparity discontinuity. The online phase is fast where a 2D image with unknown pose is matched with the closest virtual images in sampled poses. Experiments show that our approach outperforms the-state-of-the-art approaches.

A new shape based segmentation framework using statistical and variational methods

In this paper, we propose a new shape based segmentation and registration of the vertebral bodies (VBs) in clinical computed tomography (CT) images. The VB and surrounding organs have very close gray level information and there are no strong edges in some CT images. To overcome these challenges, image appearance and shape information of VBs are used. There are three phases of our experiments: i) the detection of the VB region using the Matched filter, ii) initial segmentation using the graph cuts which integrates the intensity and spatial interaction models, iii) registration of the shape priors and initially segmented region to obtain the final segmentation. Preliminary results show that our proposed algorithm gives very encouraging results and can solve many segmentation and registration problems.

Dynamic Weighting of Facial Features for Automatic Pose-Invariant Face Recognition

This paper proposes an automatic pose-invariant face recognition system. In our approach, we consider the texture information around the facial features to compute the similarity measure between the probe and gallery images. The weight of each facial feature is dynamically estimated based on its robustness to the pose of the captured image. An approach to extract the 9 facial features used to initialize the Active shape model is proposed. The approach is not dependent on the texture around the facial feature only but incorporates the information obtained about the facial feature relations. Our face recognition system is tested on common datasets in pose evaluation CMU-PIE and FERET. The results show out-performance of the state of the art automatic face recognition systems.

Rejecting pseudo-faces using the likelihood of facial features and skin

Face detection and Facial feature extraction are considered among the most studied topics in the field of biometrics. In real-world uncontrolled scenarios, high rate of false alarm is still a major problem. This paper presents a solution to reduce false alarm rate resulting from any generic face detector, through a fast post-processing algorithm based on utilizing a probabilistic framework for facial feature extraction combined with skin model. In particular, the facial features detection method inspects the geometry of the face and the likelihood of each facial feature location while the skin detector utilizes the complimentary information of color. Experimental results show a significant improvement over the state-of-the-art on the FDDB and LFPW database.

Non-Lambertian Model-based Facial Shape Recovery from Single Image Under Unknown General Illumination

Through depth perception, humans have the ability to determine distances based on a single 2D image projected on their retina, where shape-from-shading (SFS) provides a mean to mimic such a phenomenon. The goal of this paper is to recover 3D facial shape from a single image of unknown general illumination, while relaxing the non-realistic assumption of Lambert Ian reflectance. Prior shape, albedo and reflectance models from real data, which are metric in nature, are incorporated into the shape recovery framework. Adopting a frequency-space based representation of the image irradiance equation, we propose an appearance model, termed as Harmonic Projection Images, which accounts explicitly for different human skin types as well as complex illumination conditions. Assuming skin reflectance obeys Torrance-Sparrow model, we prove analytically that it can be represented by at most 5th order harmonic basis whose closed form is provided. The recovery framework is a non-iterative approach which incorporates regression-like algorithm in the minimization process. Our experiments on synthetic and real images illustrate the robustness of our appearance model vis-a-vis illumination variation.

Long term facial parts tracking in thermal imaging for uncooperative emotion recognition

This paper proposes a tracking approach for regions of interest (ROI) in thermal image videos, where vital signs can be measured for emotion recognition. The proposed tracking framework overcomes a number of problems associated with this goal; mainly size of the ROI, appearance variations in the ROI with physiological changes, and the duration of tracking in a practical setting. The proposed framework consists of three modules: An adaptive particle filter tracker, an online detector, and finally a module to integrate the outputs of the two previous modules for learning as well as the final decision. The template of the adaptive particle filter tracker is updated based on the learning decision module to avoid drifting. In the detector module, a randomized classifier is used to detect the ROI. Then the output of this classifier is enhanced by removing false positives using a proposed geometrical constraint. The proposed framework is tested and compared to the state of art approaches on 32 human subjects with different physiological changes. Experimental results show that proposed method outperforms the others.

Complex bingham distribution for facial feature detection

We present a novel method for facial feature point detection on images captured from severe uncontrolled environments based on a combination of regularized boosted classifiers and mixture of complex Bingham distributions. The complex Bingham distribution is a rotation-invariant shape representation that can handle pose, in-plane rotation and occlusion better than existing models. Additionally, we regularized a boosted classifier with a variance normalization factor to reduce false positives. Using the proposed two models, we formulate our facial features detection approach in a Bayesian framework of a maximum a-posteriori estimation. This approach allows for the inclusion of the uncertainty of the regularized boosted classifier and complex Bingham distribution. The proposed detector is tested on different datasets and results show comparable performance to the state-of-the-art with the BioID database and outperform them in uncontrolled datasets.

Passive single image-based approach for camera steering in face recognition at-a-distance applications

This paper investigates the problem of automatically steering one or more Narrow Field of View (NFOW) cameras to a target subject using only a single image from a reference NFOW camera, without the help of a Wide Field of View (WFOV) camera. To find the approximate distance of the subject from the reference camera, our algorithm uses information from facial biometrics, specifically the inter-pupil distance (IPD), and eye-to-lips distance(ELD). A trigonometric relationship is formulated to calculate the steering parameters of the other cameras. Moreover, a new robust facial feature detector is also proposed in order to estimate the required biometrics. Several experiments are reported to evaluate the proposed system.