Youssef Chahir

Nonlocal video denoising, simplification and inpainting using discrete regularization on graphs

We present nonlocal algorithms for video denoising, simplification and inpainting based on a generic framework of discrete regularization on graphs. We express video denoising, simplification and inpainting problems using the same variational formulation. The main advantage of this framework is the unification of local and nonlocal approaches for these processing procedures. We take advantage of temporal and spatial redundancies in order to produce high quality results. In this paper, we consider a video sequence as a volume rather than a sequence of frames, and employ algorithms that do not require any motion estimation. For video inpainting, we unify geometric- and texture-synthesis-based approaches. To reduce the computational effort, we propose an optimized method that is faster than the nonlocal approach, while producing equally appealing results.

Geometric and texture inpainting based on discrete regularization on graphs

We present an inpainting method for images and videos based on nonlocal discrete p-Laplace regularization on weighted graphs. Our work has the advantage of unifying local geometric methods and nonlocal exemplar-based ones in the same framework. Our image inpainting benefits from local and nonlocal regularities within the image. In addition to that, our video inpainting exploits temporal and spatial redundancies in order to obtain high quality results by considering a video sequence as a volume and not as a sequence of still frames. However, our method does not employ any motion estimation for video inpainting. Experiments demonstrate that our nonlocal method outperforms the local one by completing missing data with finer and more consistent details for textured and non-textured images and videos.

Searching Images on the Basis of Color Homogeneous Objects and their Spatial Relationship

In this paper we introduce several techniques which characterize color homogeneous objects and their spatial relationships for a more precise and efficient content-based image searching. We first present a region growing technique for efficient color homogeneous object segmentation, and then we extend the 2-D string to express spatial signatures for an accurate description of spatial relationships of objects within an image. Several optimizations, including dominant color histogram clustering, have also been proposed for an efficient search engine implementation. The experimental results that we have drawn so far show that our content-based image searching techniques give a high precision while maintaining a very good recall rate.

Moving object Segmentation; using optical flow with active contour model

The paper presents an object segmentation approach that combines optical flow and active contour model to characterize objects and follow them in video sequences. Our aims is to discriminate moving objects from a static background. The approach is based on a minimization of a functional of energy (E) which uses perceptual information in regions of interest (ROI) in an image, in conjunction with a mixture of Gaussian to model voxels of the background image and those of the visual objects. In this work, we compute the optical flow then we use the result of the optical flow as an input in an active contour model. Experiments with a number of test sequences are promising and extend the numerous works on this subject.

Video denoising via discrete regularization on graphs

We present local and nonlocal algorithms for video denoising based on discrete regularization on graphs. The main difference between video and image denoising is the temporal redundancy in video sequences. Recent works in the literature showed that motion compensation is counter-productive for video denoising. Our algorithms do not require any motion estimation. In this paper, we consider a video sequence as a volume and not as a sequence of frames. Hence, we combine the contribution of temporal and spatial redundancies in order to obtain high quality results for videos. To enhance the denoising quality, we develop a nonlocal method that benefits from local and nonlocal regularities within the video. Experiments show that the nonlocal method outperforms the local one by preserving finer details at the expense of an increase in the computational effort. We propose an optimized method that is faster than the nonlocal approach, while producing equally attractive results.

3D-Posture Recognition Using Joint Angle Representation

This paper presents an approach for action recognition performed by human using the joint angles from skeleton information. Unlike classical approaches that focus on the body silhouette, our approach uses body joint angles estimated directly from time-series skeleton sequences captured by depth sensor. In this context, 3D joint locations of skeletal data are initially processed. Furthermore, the 3D locations computed from the sequences of actions are described as the angles features. In order to generate prototypes of actions poses, joint features are quantized into posture visual words. The temporal transitions of the visual words are encoded as symbols for a Hidden Markov Model (HMM). Each action is trained through the HMM using the visual words symbols, following, all the trained HMM are used for action recognition.

Video Denoising and Simplification Via Discrete Regularization on Graphs

In this paper, we present local and nonlocal algorithms for video denoising and simplification based on discrete regularization on graphs. The main difference between video and image denoising is the temporal redundancy in video sequences. Recent works in the literature showed that motion compensation is counter-productive for video denoising. Our algorithms do not require any motion estimation. In this paper, we consider a video sequence as a volume and not as a sequence of frames. Hence, we combine the contribution of temporal and spatial redundancies in order to obtain high quality results for videos. To enhance the denoising quality, we develop a robust method that benefits from local and nonlocal regularities within the video. We propose an optimized method that is faster than the nonlocal approach, while producing equally attractive results. The experimental results show the efficiency of our algorithms in terms of both Peak Signal to Noise Ratio and subjective visual quality.

Application of BSIF, Log-Gabor and mRMR transforms for iris and palmprint based Bi-modal identification system

Verification of individual identity through the process of biometric identification involves comparison between an encoded value and a stored value of the biometric feature in question. The effectiveness of a multimodal user authentication system is greater, but so is its complexity. The system error rate is reduced by the fact that multiple biometric features are combined, thus solving the weakness of the single biometric. Performance of individual authentication through palm-print- and iris-based bimodal biometric system is proposed in the present study. To this end, Log-Gabor filter and BSIF (Binarised Statistical Image Feature) coefficients are employed to obtain the iris and palm-print traits, and subsequently selection of the features vector is conducted with mRMR (Minimum Redundancy Maximum Relevance) transforms in higher coefficients. To match the iris or palm-print feature vector, the Hamming Distance is applied. According to the experiment outcomes, the proposed system not only has a significantly high recognition rate but it also affords greater security compared to the single biometric system.