F23: Geodesic Stripes Based Hierarchical Evaluation for 3D Facial Similarity

Hongyan Li,zhongke Wu, Donghua Huang

Similarity evaluation of 3D face is the core issue in 3D face recognition. The article puts forwards a geodesic stripes based evaluation method which realizes assessment from global face to local parts. It can be used to make effective and comprehensive evaluation in many fields such as 3D face reconstruction, forensic science, archaeology etc. First, the method applies nose tip overlap to eliminate translation difference and the classical PCA and ICP alignment algorithm to eliminate rotation difference. Then simplify each 3D face with a series of geodesic stripes and calculate distribution vector between each pair of stripes, which reflects 3D space distribution feature. Finally through feature extraction on entire face, we get a distribution matrix which consists of all distribution vectors. The similarity between two distribution matrices directly shows the global similarity between two faces. We also extract geodesic stripes feature on local organs like mouth eye and nose to make a more accurate evaluation. The experimental results on SHREC2008 3D face database further testify that the hierarchical evaluation method is available and consistent with the subjective evaluation.