Part-to-full shape matching of different human subjects

Shape matching is a fundamental operation in digital geometry processing and computer graphics. Challenges in shape matching include finding correspondences of partial shapes with deformations, as well as topological noise and ambiguities. This paper presents a partial shape correspondence algorithm based on the concept of the functional map. An iterative dense matching algorithm, incorporating sparse and guided dense matching, is proposed along with a new objective function including both descriptor matching error and transformation error. Rank estimation with the rank direction is proposed to achieve more accurate slope approximation of the functional map. The slope is beneficial because it directly influences the matching efficiency. The experimental results obtained using FAUST and SHREC'16 datasets demonstrate the effectiveness of our proposed algorithm for matching the shapes of different human subjects and shapes with large missing parts compared with state-of-the art algorithms. The proposed algorithm provides an average geodesic distance of <0.033 even when the missing part is up to 80% of the area.