Three-polar parameterization-based ICP algorithm for 3D face registration and recognition

Majdi Jribi, Arwa Habbachi, Faouzi Ghorbel

Abstract: We propose, in this work, a novel registration-based approach for the 3D face recognition task. A new version of the well known Iterative Closest Point (ICP) algorithm is defined, in this context. It is performed using the three-polar parameterization implemented on 3D faces as input to the ICP method instead of using the whole face point cloud. The proposed approach establishes a faster and more robust version of the ICP algorithm than the classical one since the three-polar parameterization permits to obtain an efficient and ordered set of points. The performances of the proposed approach are tested on the BU-3DFE and Bosphorus databases after applying random rigid transformations on each face. The obtained results show that the novel version of the ICP algorithm outperforms the classical version in the registration procedure and, thus, in the face recognition operation. The obtained rates of recognition are very competitive with the state of the art methods.

External IDs:dblp:journals/sivp/JribiHG25