We propose a novel grayness index for finding gray pixels and demonstrate its effectiveness and efficiency in illumination estimation. The grayness index, GI in short, is derived using the Dichromatic Reflection Model and is learning-free. The proposed GI allows estimating one or multiple illumination sources in color-biased images. On standard single-illumination and multiple-illumination estimation benchmarks, GI outperforms state-of-the-art statistical methods and many recent deep net methods. GI is simple and fast, written in a few dozen lines, processing a 1080p image in about 0.4 seconds with a non-optimized Matlab code.