Abstract
Estimation of scene illumination from a single image or an image sequence has been widely studied in computer vision. The approach presented in this paper, introduces two new issues: (1) illumination classification is performed rather than illumination estimation; (2) an object-based approach is used for illumination evaluation. Thus, pixels associated with an object are considered in the illumination estimation process using the object's spectral characteristics. Simulation and real image experiments, show that the object-based approach indeed improves performance over standard illumination classification.
Original language | English |
---|---|
Pages (from-to) | 1723-1732 |
Number of pages | 10 |
Journal | Pattern Recognition |
Volume | 35 |
Issue number | 8 |
DOIs | |
State | Published - Aug 2002 |
Keywords
- Classification
- Color
- Color correction
- Face
- Illumination estimation
ASJC Scopus subject areas
- Software
- Signal Processing
- Computer Vision and Pattern Recognition
- Artificial Intelligence