Computational color constancy is a classical problem in computer

vision. It is an under-constrained problem, which can be solved

based on some constraint.  Existing algorithms can be divided into

two groups:physics-based algorithms and statistics-based

approaches. In this paper, we propose a new hypothesis that the

images generated under a same illumination have some similar

features. Based on this hypothesis, a novel statistics-based color

constancy algorithm is given and a new similarity function between

images is also defined. The experimental results show that our

algorithm is effective and it is more important that the dimension

of the features in our algorithm is much lower than many former

statistics-based algorithms.