Rachid Benslimane

Fibre width measurement and quantification of filler size distribution in paper-based materials by SEM and image analysis

In this paper the development of image processing and analysis routines for acquiring information about the paper structure from images obtained by scanning electron microscopy (SEM) is described. A new method to measure the fibre width based on the processing of superficial SEM observation is herein proposed and successfully applied to some commercial paper-based materials. The quantification of fillers in the Z-direction of paper was based on paper cross-sectional SEM images. The results obtained by image processing demonstrated the excellent performance and suitability of the proposed method to quantify exactly information concerning fraction and size distribution of digital fillers. The fraction of digital fillers was correlated with the surface roughness of a given paper grade.

Symmetrical motif extraction for periodic ornamental patterns

This paper presents an automatic method to extract symmetrical motif for periodic ornamental patterns. It is well known that for every periodic pattern a lattice of points is associated. The points of this lattice are inter-related by two translational vectors, which define the translational unit representing the repetitive motif. This repetitive motif must be Symmetric if we are interested to identify the symmetry group to which a particular pattern belongs. To do so, we propose a new method based on autocorrelation function ACF and SIFT algorithm. Tests on 166 images characterized with different visual quality demonstrate the capability of the proposed method for automatic periodic motif extraction without the need of human intervention.

A Genetic Based Algorithm for Automatic Motif Detection of Periodic Patterns

In this paper, we introduce a new method to extract the basic unit cell of a periodic Islamic Geometrical Pattern (IGP). This method is based on the autocorrelation function (ACF), a function known to be appropriate to analyze and draw out a repetitive motif from a regular texture. The motif can be successfully extracted when the detected peaks in the autocorrelation function of an image are pertinent. In this paper, the pertinent peaks detection is considered as an optimization problem. For this purpose, we propose a criterion function that we optimize by using genetic algorithm. The proposed method is tolerant to geometric distortion, and photometric quality of the input image compared to classical ones. Tests on 166 images with different visual quality demonstrate the capability of the proposed method to extract the periodic motif automatically.

Arabo-Moresque decor image retrieval system based on mosaic representations

Abstract The paper describes a new method for indexing an Arabo-Moresque decor database. This method requires that the images in the database must have the whole principal geometric information, named ‘spine’ of the decor. In practice, it is difficult to respect this request. When the distance between the camera and the real scene (fresco in Zellij) is big, the whole spine is captured but the resolution of the image is bad. On the other hand, when this distance is small, the resolution is high but the spine is not completely captured. This motivates the use of the ‘mosaicing’ technique. The contribution of this work is the combination of the mosaicing and indexing techniques for the development of an Arabo-Moresque decor image retrieval system.

System of indexing, annotation and search in the old Arabic manuscripts

Ancient manuscripts constitute a heritage to preserve for future generations and make it accessible to a wider audience. This heritage is unfortunately consulted by a minority of experts and researchers because of the lack of information on the existence of this national heritage. The problem is partly due to the fact that this manuscript heritage is stocked in libraries scattered around the country and managed by institutions belonging to different departments or different establishments of the same department. On the other hand, restriction of access to national heritage manuscript is related to the concern to preserve the manuscripts physically manipulated which contribute to their accelerated degradation, taking into consideration these limitations on access while ensuring preservation of original manuscripts, the solution widely adopted by developed countries is based partly on the digitization of this heritage manuscript, and partly on the development of management platforms and diffusion of this wealth of knowledge digitized. Easy access to such manuscripts presented on images format requires an index which can be created manually or automatically by using Optical Recognition Characters (OCR). The automatic approach is difficult to realize by considering the cursive nature of Arabic writing and by the exaggerated overlapping between words and lines in handwritten Arabic manuscripts. So, the segmentation leading to extract respectively and separately lines, words and characters constitutes the critical operation, affecting the performance of all Arabic (OCR) systems. Recent research works adopt the word spotting approach. The words in a manuscript are matched as images and grouped into clusters which contain all instances of the same word. This new and original approach solve the problem related to the cursive nature of the Arabic writing by not considering the character segmentation, but considers the line and word segmentation perfectly performed. To tackle these problems mainly connected to segmentation operation, we propose in this paper a new indexing system of Arabic manuscripts, which facilitates transcription and the establishment of Arabic handwritten text by annotating images of manuscripts according to metadata. Our system offers the ability to create works of Arabic manuscripts in a format TEI XML, this model includes all the primitives that are related to the international standard markup TEI (Text Encoding Initiative), and to provide good guarantees compatibility with existing tools. This system supplies the database of the open source platform SDX (System Documentary XML) that allows searching, browsing and viewing of these documents in a web format. The good performances of the proposed method were tested by using some ancient Arabic manuscripts.

A genetic algorithm for recovering camera parameters and motion from silhouettes

We present a novel approach for recovering the camera parameters and motion from the concept of coherence of a set of silhouettes of an object taken under circular motion. These parameters can be obtained by maximizing the total coherence between all silhouettes. For this purpose, we propose in this paper to use genetic algorithm instead the conventional optimization procedures. Genetic algorithm can correctly find the optimal solution without the need of initial values of parameters. The performance of the proposed method is shown in term of convergence and accuracy to recover camera parameters. The recovered parameters may be used for 3D image-based modeling to obtain high quality 3D reconstructions.