To observe the changing rule of clustering center’s value in K-means by the YCbCr color space, against a lot of cell image by Wright Stain, this paper proposed a new method to determine the number of clusters in K-means to get a good result of color correction. Different target of cell image could be firstly gathered in the corresponding class accurately, then by matching between classes, and using histogram specification for adjusting histogram, the results of color correction could at last be achieved. It has been proved by many experiments that this algorithm can greatly improve the accuracy of color correction results, especially the target with closer color features in cell.