COMPARISON METHODS OF STEREO VISION ALGORITHMS FOR 3D FACE RECONSTRUCTION
Read the full article
The results of comparative analysis of the stereo algorithms effectiveness for solving the problem of 3D model of human face reconstruction are presented. The methods of stereo vision relating to the three main classes – local, global and semi-global methods, are considered. The measure of differences between recovered depth map and the reference depth map obtained by a camera with a structured infrared illumination is proposed as the criteria for quantitative evaluation of the algorithms. During the comparison, transition to 3D space was carried out, calculation of the characteristics for the point cloud was performed, the quality of 3D surface restored with a depth map and performance of algorithms were considered. Selection of the best algorithm based on quantitative and qualitative evaluation of the algorithms was performed. The results can be used to improve the efficiency of stereo vision methods for identification of a person.