Roaa K H A L I D SOLOH | Université du Havre (original) (raw)

Ph.D. candidate at Lebanese University- Le Havre University.Works on graphs and optimization modelling for matching problems
Address: Beirut, Beyrouth, Lebanon

less

Uploads

Papers by Roaa K H A L I D SOLOH

The widespread of 3D shapes nowadays, gained it huge importance in several fields like computer v... more The widespread of 3D shapes nowadays, gained it huge importance in several fields like computer vision, engineering, image processing, and many others. Its main challenge is the representation of these shapes and projecting them into canonical features referred to as descriptors. The necessity of them appears in different tasks like classification, retrieval, and matching, where they considered as the main step in what follows. Moreover, the matching problem is the core of all other tasks. This is why in this paper we propose a graph matching problem to find a one-to-one correspondence between models, it's obviously known as the NPhard problem. So, a novel compact feature vector to represent our 3D models is extracted, combining several geometric representative curvatures, it's simple in complexity computational, yet powerful discriminating in the sense of affine transformations. The 3D surface is modeled as an undirected weighted graph, with the Gaussian kernel as a weight function. Integer Linear Programming is used in order to segment our meshes into regions, where we maximized the modularity between vertices, these regions are represented by a single

The widespread of 3D shapes nowadays, gained it huge importance in several fields like computer v... more The widespread of 3D shapes nowadays, gained it huge importance in several fields like computer vision, engineering, image processing, and many others. Its main challenge is the representation of these shapes and projecting them into canonical features referred to as descriptors. The necessity of them appears in different tasks like classification, retrieval, and matching, where they considered as the main step in what follows. Moreover, the matching problem is the core of all other tasks. This is why in this paper we propose a graph matching problem to find a one-to-one correspondence between models, it's obviously known as the NPhard problem. So, a novel compact feature vector to represent our 3D models is extracted, combining several geometric representative curvatures, it's simple in complexity computational, yet powerful discriminating in the sense of affine transformations. The 3D surface is modeled as an undirected weighted graph, with the Gaussian kernel as a weight function. Integer Linear Programming is used in order to segment our meshes into regions, where we maximized the modularity between vertices, these regions are represented by a single

Log In