Abstract | ||
---|---|---|
We have proposed a decomposition of the eigen- face into two orthogonal eigenspaces and have shown that the decomposition is effective for realizing robust face recognition under various lighting conditions (lo). The present paper refines the decomposed eigenface method by introducing a projection-based image cor- rection. The image correction technique is principally authorized when the object shape is fixed and a suf- ficient number of images are taken beforehand. How- ever, the proposed technique can also be applied to a canonical eigenspace, which is constructed from several faces taken under various lighting conditions. Reflec- tive noises, shadows and occlusions are detected and corrected by the projection of a facial image onto the canonical eigenface. Based on the newly proposed im- age correction, we develop herein a refined decomposed eigenface method. The experimental results indicate that the refinement works well for face recognition un- der various lighting conditions, as compared to the original decomposed eigenface method. |
Year | Venue | Keywords |
---|---|---|
2002 | MVA | face recognition |
Field | DocType | Citations |
Computer vision,Facial recognition system,Image correction,Eigenface,Pattern recognition,Artificial intelligence,Mathematics,Eigenvalues and eigenvectors | Conference | 0 |
PageRank | References | Authors |
0.34 | 4 | 3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Kazuma Shigenari | 1 | 3 | 1.07 |
Fumihiko Sakaue | 2 | 46 | 19.84 |
Takeshi Shakunaga | 3 | 192 | 43.46 |