View influence analysis and optimization for multiview face recognition
Journal on Image and Video Processing
| Hi-index | 0.00 |
Many face database and recognition systems havebeen constructed in specially designed studios withvarious illuminations, poses, and expressions.However, none of these databases yet satisfies a largevariation of poses and illuminations that permit thestudy of systematic 3D human face information, whichresults in an unsatisfactory success rate. It is causedprimarily by the difficulty of data collection of facialimages to satisfy the large variation of poses andilluminations to fully represent the 3D characteristicsof human faces. We present how computer-generateddatabase can be used for face recognition experimentfocused on multi-view face recognition/descriptor. Weshow multi-view face data collection using renderingof 3D models. We also illustrate our approach how tobuild a face descriptor containing 3D information ofhuman face using multi-view concepts. This multi-viewface recognition descriptor is a 3D descriptor usingthe concept how much powerful a view influences overnearby views, so called as "quasi-view" size.