This paper presents our method for reconstructing facial image from a partially occluded facial image or a low-resolution one using example-based learning. Faces are modeled by linear combinations of prototypes of shape and texture. With the shape and texture information from an input facial image, we can estimate optimal coefficients for linear combinations of prototypes of shape and texture by simple projection for least square minimization. The encouraging results of the proposed method show that our method can be used to improve the performance of the face recognition by reconstructing facial image from a partially occluded facial image or a low-resolution one.
|Number of pages||14|
|Journal||Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)|
|Publication status||Published - 2004|
ASJC Scopus subject areas
- Theoretical Computer Science
- Computer Science(all)