Efficient two-level image thresholding method based on Bayesian formulation and the maximum entropy principle

An efficient method for two-level thresholding is proposed based on the Bayes formula and the maximum entropy principle, in which no assumptions of the image histogram are made. An alternative criterion is derived based on maximizing entropy and used for speeding up the searching algorithm. Five forms of conditional probability distributions-simple, linear, parabola concave, parabola convex, and S-function-are employed and compared to each other for optimal threshold determination. The effect of precision on optimal threshold determination is discussed and a trade-off precision ε=0.001 is selected experimentally. Our experiments demonstrate that the proposed method achieves a significant improvement in speed from 26 to 57 times faster than the exhaustive search method.