Abstract: An algorithm for estimating illuminant chromaticity is proposed and highlight is removed based on specular-to-diffuse mechanism. First, best fit window (BFW) is introduced. Pixels in BFW then are normalized and projected into 2D inverse-intensity space. Then, fuzzy C-mean clustering algorithm is performed to cluster the data in inverse-intensity space into three clusters. Finally, linear fitting approach is implemented based on two clusters to obtain intercept value as the estimated illuminant chromaticity. Our algorithm is simple and easy in implementation. Experiments show that the method is highly efficient.