Binarization algorithm based on differential evolution algorithm for gray images

To solve the image binarization, an image binarization algorithm based on the differential evolution algorithm, called BAbDE, is presents in this paper. BAbDE randomly generates a population of 2-dimention vectors (individuals) as the set of initial centers, BAbDE applies the spirit of `survival of the fittest´ implied in the classical differential evolution algorithm to obtain the optimal binary image of an image. Numerical experiments are firstly conducted to study the setting of two control parameters, the mutation factor and the crossover probability, of BAbDE. BAbDE is then compared with Otsu´s method and K-means method in their thresholds, processing times and entropies. The experimental results for several common-used images show that BAbDE is practicable for image binarization, and its entropy is smaller than the ones of another two methods.