Abstract: This paper points out that multilevel thresholds determined by Kapur's method depend on each other.Based on the property, a fast implementation of Kapur's method for multilevel thresholding is proposed.With an initial thresholding value, the proposed method takes an iterative way to compute the multilevel thresholds until the convergent condition is met.Experimental results show that the proposed method outperforms many other methods for both runtime reduction and accuracy.Furthermore, the proposed method is quite simple and has a less requirement of memory which makes it meet the requirement of real-time.