Research of improved genetic algorithm for thresholding image segmentation based on maximum entropy

In the paper a novel improved genetic algorithm is proposed based on the maximum entropy for thresholding image segmentation. First of all, the encoded mode is made and the maximum entropy function is selected as the key adaptation genetic algorithm, and then the initial group is generated by roulette selection algorithm to the next generation for the best individual, which can improve the global search capability of genetic algorithm, crossover probability and mutation probability. Comparing with the standard genetic algorithm, the improved algorithm reduce times of computing and enhance efficiency of computing. Form the results of the experiment we can see that the improved algorithm has also some advantages, such as quickly, validity and practicability.