Aiming at the problem of threshold selection in image segmentation, a multi threshold segmentation algorithm based on grasshopper algorithm is proposed in this paper. In this algorithm the otsu method and Kapur’s entropy are considered. The fitness function which used are the maximum between class variance criterion (Otsu) and the Kapur’s Entropy. This method uses the grasshopper optimization algorithm to optimize threshold. In the end, the image is segmented with the best threshold. The algorithm is compared with the traditional Otsu algorithm, the maximum entropy method, the image segmentation method based on particle swarm, and the image segmentation method based on artificial bee colony. Experimental results show that the algorithm is better than other algorithms. When the number of thresholds is 4 and 5, the PSNR of the proposed algorithm is about 3% and 15% higher than that of particle swarm optimization algorithm and artificial bee colony algorithm. The running time of this algorithm is about 9% and 5% faster than that of particle swarm optimization algorithm and artificial bee colony algorithm.