Advanced human body and head shape representation and analysis
ICDHM'07 Proceedings of the 1st international conference on Digital human modeling
3D face recognition based on geometrical measurement
SINOBIOMETRICS'04 Proceedings of the 5th Chinese conference on Advances in Biometric Person Authentication
Consensus strategy for clustering using RC-images
Pattern Recognition
Hi-index | 0.01 |
The importance of fit for face-related wearing products has introduced the necessity for better definition of face area. In this paper, three definitions of face area are compared on the context of Three dimensional (3D) face shape similarity based clustering. The first method defines the face area by spanning from the whole head grid surface by the front *** /2 wedge angle along a line going through the centroid and pointing to the top of the head. The second method defines the face area as the grid surface enclosed by several anthropometric landmark points (sellion, both zygions, and menton) on the facial surface. The zonal surface where the respirator interferes with the wear's face is taken as the third alternative definition for the comparative study. By utilizing the block-distance measure, each face was converted into a compact block-distance vector. Then, k-means clustering was performed on the vectors. 376 3D face data sets were tested in this study. One-way ANOVA on the block distance based vectors was conducted to evaluate the influence on clustering results by utilizing different face area definitions. No difference was found at the significant level of 0.05. However, the cluster membership shows great difference between different definitions. This emphasizes the value of the selection of face area in 3D face shape-similarity-based clustering.