| Budget Amount *help |
¥2,500,000 (Direct Cost: ¥2,500,000)
Fiscal Year 1997: ¥1,100,000 (Direct Cost: ¥1,100,000)
Fiscal Year 1996: ¥1,400,000 (Direct Cost: ¥1,400,000)
|
|---|
| Research Abstract |
Surface curvatures such as Gaussian, mean and principal curvatures are intrinsic surface properties and have playd important roles in curved surface analysis. However, only their signs are reliably used in classifying and identifying surfaces.
In this paper, we treat face recognition problem as 3D shape recognition problem of free-form curved surfaces and present an approach based on the analysis of two principal curvatures, max and minimum curvatures and their directions. The approach is based on simple correlation-based matching algorithm which does not require either face feature extraction or surface segmentation.
Each face in both input images and the model database, is represented as an Extended Gaussian Image(EGI), constructed by mapping principal (max and minimum) curvatures and their directions onto the unit sphere, estimated at each surface points. Individual face is then recognized by evaluating the similarities among others by using Fisher's spherical correlation.
The method is tested for its simplicity and robustness and successively implemented for each of 37 face range images from NRCC (National Research Council Canada) 3D image data files. It turned out that the method is quite simple, efficient and robust to distractions such glasses, facial hair or hair style. These represent important advantages over conventional method based 2D features such as eye, mouth patterns. Results shows that shape information from surface curvatures provides vital cues in distinguishing and identifying such fine surface structure as human faces.
|
