Slimane Larabi

User-independent system for sign language finger spelling recognition

Abstract We propose in this paper a framework for recognizing the sign language alphabet. To separate hand images from complex backgrounds, we use skin colour and texture attributes with neural networks. The recognition process is based on the combination of three shape descriptors: Discrete orthogonal Tchebichef moments applied on both internal and external outlines hand, Hu moments and a set of geometric features derived from the convex hull that encloses the hand shape taking into account the hand orientation. The recognition is carried out using KNN and SVM classifiers. The proposed descriptors are combined in several sequential and parallel manners and applied on different datasets. The obtained results are compared to existing works.

LWDOS: language for writing descriptors of outline shapes

A new method to obtain a rough description of a 3-D object from its outline shape is presented in this paper. Firstly the outline shape is split up into parts that are related in a connectivity graph. The boundary features of every part (lines, curves) and the information about its junction with other parts are extracted in order to provide a semantic content to the outline shape graph. A specific language to describe this graph has been developed. Finally, some results obtained applying our approach over real images are presented and discussed.

Background Subtraction Algorithms with Post-processing: A Review

Due to its several algorithms with their fast implementations, background subtraction becomes a very important step in many computer vision and video surveillance systems which assume static cameras. Literature counts a large number of robust background subtraction algorithms which try each to outperform the others in a quantitative and qualitative manner. This competition can sometimes confuse the user of this kind of process and make the choice of one of them difficult. To overcome this issue we review, in what follows, the background subtraction process by defining it and exploring most used algorithms of background subtraction. We then expose some post processing techniques used to remove superfluous content derived from background subtraction.

Accurate real-time neural disparity MAP estimation with FPGA

We propose in this paper a new method for real-time dense disparity map computing using a stereo pair of rectified images. Based on the neural network and Disparity Space Image (DSI) data structure, the disparity map computing consists of two main steps: initial disparity map estimation by combining the neuronal network and the DSI structure, and its refinement. Four improvements are introduced so that an accurate and fast result will be reached. The first one concerns the proposition of a new strategy in order to optimize the computation time of the initial disparity map. In the second one, a specific treatment is proposed in order to obtain more accurate disparity for the neighboring pixels to boundaries. The third one, it concerns the pixel similarity measure for matching score computation and it consists of using in addition to the traditional pixel intensities, the magnitude and orientation of the gradients providing more accuracy. Finally, the processing time of the method has been decreased consequently to our implementation of some critical steps on FPGAs. Experimental results on real datasets are conducted and a comparative evaluation of the obtained results relative to the state-of-art methods is presented.

Curve normalization for shape retrieval

In this paper, we propose a novel part-based approach for two dimensional (2-D) shape description and recognition. According to this method, first the polygonal approximation is employed to represent the outline shape by an ordered sequence of parts. Then using the Least squares model, each part is associated with a cubic polynomial curve. The obtained curves are normalized that are invariant to scaling, rotation and translation. Finally, based on shape similarity of resulting curves, a shape similarity between an input shape and its reference model is defined. A two-step matching algorithm is proposed. Experiments using several benchmark databases are performed and the obtained retrieval results demonstrate that the proposed approach is effective as compared to other matching techniques.

Head pose estimation based on face symmetry analysis

This paper addresses the problem of head pose estimation in order to infer non-intrusive feedback from users about gaze attention. The proposed approach exploits the bilateral symmetry of the face. Size and orientation of the symmetrical area of the face is used to estimate roll and yaw poses by the mean of decision tree model. The approach does not need the location of interest points on face and presents robustness to partial occlusions. Tests were performed on different datasets (FacePix, CMU PIE, Boston University) and our approach coped with variability in illumination and expressions. Results demonstrate that the changes in the size of the regions that contain a bilateral symmetry provide accurate pose estimation.

Indexing binary images using quad-tree decomposition

The tree structure is introduced to specify block-oriented decomposition of database images. These decomposition structures offer a fundamental data model for specifying image content in large image databases.

Obstacle Detection from Uncalibrated Cameras

In this paper, we propose a new approach for obstacle detection based on the analysis of images taken by uncalibrated stereo rig. This system can be divided in to two main stages. The first one deals with computing the fundamental matrix from the matching between points of interest in order to compute a dense disparity map. Whereas the second one presents a very simple and faster method for obstacle detection, by using the segmentation image. Indeed, the combination of the segmented image and the disparity map are going to permit us to extract the vertical 3D segments that will indicate us the presence of obstacles in the scene. This approach allows us to detect several numbers of obstacles of varied shapes and sizes. This obstacle detection stage can be viewed as the first stage of a free space estimator which can be implemented in a mobile robot.

Efficient descriptor for full and partial shape matching

In this paper we present a new approach for full and partial shape retrieval based on a shape descriptor invariant to geometric transformations, reflection and deformation. The proposed description is a set of features that capture simultaneously global and local properties of the shape. To achieve the best matching, we propose a novel matching algorithm based on Dynamic Time Warping. The proposed method is evaluated in two cases: partial and full matching. The experimental results demonstrate that our approach outperforms existing methods of partial shape retrieval and gives comparable results for full shape retrieval.

Inferring heading direction from silhouettes

Due to the absence of features that may be extracted from face, heading direction estimation for low resolution images is a difficult task. For such images, estimating heading direction requires to taking into account all information that may be inferred from human body in image, particularly its silhouette. We propose in this paper a set of geometric features extracted from shape shoulders-head, feet and knees shapes which jointly allow the estimation of body direction. Other features extracted from head-shoulders are proposed for the estimation of heading direction based on body direction. The constraint of camera position related to proposed features is discussed and results of experiments conducted are presented.