Representing and analyzing relief patterns using LBP variants on mesh manifold
- PDF / 5,271,748 Bytes
- 17 Pages / 595.276 x 790.866 pts Page_size
- 99 Downloads / 178 Views
THEORETICAL ADVANCES
Representing and analyzing relief patterns using LBP variants on mesh manifold Claudio Tortorici1 · Naoufel Werghi2 · Stefano Berretti3 Received: 20 March 2019 / Accepted: 10 September 2020 © Springer-Verlag London Ltd., part of Springer Nature 2020
Abstract Extending the concept of texture to the geometry of a mesh manifold surface is an emerging topic in computer vision. This concept is different from gluing images to the surface, but rather indicates the presence of relief patterns that locally change the surface geometry, showing some regular and repetitive patterns. The representation and the analysis of such relief patterns have several potential applications. In this paper, we propose an original and comprehensive framework to address this novel task, which redefines a large variety of local binary patterns on the mesh manifold domain. We also propose an efficient mesh re-sampling technique that enables uniform surface tessellation. We assess the different descriptive variants derived with this framework in terms of uniformity, repeatability and discriminative power. Afterward, we conduct an extensive experimentation on different datasets showcasing the competitiveness of our framework in classification and retrieval tasks, in terms of both accuracy and computational complexity, with respect to state-of-the-art methods. Keywords Relief patterns representation and classification · Mesh-LBP variants · Ordered ring facets · Mesh regularization
1 Introduction Using mesh manifolds as input data, several studies addressed the problem of retrieving/classifying 3D shapes based on their similarities [6, 31]. In most of the cases, synthetic models generated by ad-hoc software have been used, while reconstructed meshes have been considered more rarely. An even less investigated, but emerging problem, which is of interest for its potential application in several contexts is the classification of 3D relief patterns. As peculiar trait, the style of these patterns does not depend on the overall structure of the shape. They are rather characterized by some form of regularity and repeatability across the surface so that they can be regarded as the 3D geometric * Claudio Tortorici [email protected] Naoufel Werghi [email protected] Stefano Berretti [email protected] 1
Technology Innovation Institute, Abu Dhabi, UAE
2
Khalifa University, Abu Dhabi, UAE
3
University of Florence, Florence, Italy
equivalent of textures in 2D images. Examples are knitted fabrics [8], artworks patterns, artists styles or natural structures like tree barks [26], rock types or engravings [43], etc. Modeling and analyzing such bas-reliefs from 3D objects revealed attractive in modern Computer Vision [2, 13, 30, 33]. Figure 1 shows some examples of relief patterns as acquired in different domains, i.e., medical, archaeological, geographical. The availability of datasets that include relief patterns makes now of increasing relevance the problems of detecting, retrieving and classifying them.
Data Loading...
