A novel method for automatic color grading of beef lean tissue was developed using computer vision and BP neural network (BP-NN) techniques. 160 beef rib-eye cross-section images were collected and color score of each sample were manually determined by a five-number panel. The segmentation of lean tissue region from rib-eye cross-section image was carried out and color features of each image were extracted using computer image processing technologies. A BP neural network model, with inputs of color features and outputs of color scores, respectively, was designed to automatically estimate the grade of beef lean tissue color. The optimum structure parameters of the BP model were determined by training. Finally, the proposed BP model was employed to predict the color score of each sample in validation set. Results show that average 95% of samples can be assigned a correct color score in 0.25s, indicating that the proposed method permits accurate, rapid and reliable prediction of color grade of beef lean tissue.