Aiming at removing complex soil background noise in the corn seedling filed, a color image segmentation algorithm based on MMC (Maximum margin criterion) and CV (Chan-Vese) was proposed. The corn color image was transformed into gray image by using MMC, and the grayscale image was denoised by TV (Total variation) filter. Then filtered image was segmented by the CV model. The results of the experiment by Matlab showed that the algorithm could effectively get the extraction of the objection of corn and noise reduction of weed and moss simultaneously in the image. The misclassification rate and the leakage rate were 4.32% and 9.69% respectively, and the similarity was 86.57%.