Illumination Estimation via Non-Negative Matrix Factorization

Shi, L., Funt, B., Xiong, W., Kim, S., Kang, B., and Lee, S.D., "Illumination Estimation via Non-negative Matrix Factorization," Proc. AIC 2007 Color for Science and Industry, Midterm Meeting of the International Color Association , Hangzhou, July 2007.


The problem of illumination estimation for colour constancy and automatic white balancing of digital color imagery can be viewed as the separation of the image into illumination and reflectance components. We propose using nonnegative matrix factorization with sparseness constraints (NMFsc) to separate the components. Since illumination and reflectance are combined multiplicatively, the first step is to move to the logarithm domain so that the components are additive. The image data is then organized as a matrix to be factored into nonnegative components. Sparseness constraints imposed on the resulting factors help distinguish illumination from reflectance. Experiments on a large set of real images demonstrate accuracy that is competitive with other illumination-estimation algorithms. One advantage of the NMFsc approach is that, unlike statistics- or learning-based approaches, it requires no calibration or training.

Full text (pdf)


Back to SFU Computational Vision Lab publications (home)