Multilevel thresholding for image segmentation has always been a popular issue and has attracted much attention. Traditional exhaustive search methods take considerable time to solve multilevel thresholding problems. However, heuristic search algorithms have potential advantages in terms of solving such multilevel thresholding problems. Based on this idea, in this paper, a novel adaptive gravitational search algorithm (AGSA) is proposed to solve the optimal multilevel image thresholding problem; this algorithm is more efficient than the traditional exhaustive search method for grayscale image segmentation. In the AGSA, an adaptive parameter optimization strategy is used to tune the gravitational constant and the inertia weight. To verify the performance of the proposed algorithm, a series of classic test images are used to perform several experiments. In addition, the standard GSA and some optimization algorithms are compared with the proposed algorithm. The experimental results show that the proposed algorithm is obviously better than the other six algorithms. These promising results suggest that the AGSA is more suitable than existing methods for multilevel image thresholding.

用于图像分割的多级阈值一直是一个流行的问题，并引起了广泛的关注。传统的穷举搜索方法需要花费大量时间来解决多级阈值问题。但是，启发式搜索算法在解决此类多级阈值问题方面具有潜在的优势。基于这种思想，本文提出了一种新的自适应重力搜索算法（AGSA）来解决最优的多级图像阈值问题。与传统的穷举搜索方法进行灰度图像分割相比，该算法效率更高。在AGSA中，使用自适应参数优化策略来调整重力常数和惯性权重。为了验证所提出算法的性能，一系列经典的测试图像用于执行多个实验。此外，将标准GSA和一些优化算法与所提出的算法进行了比较。实验结果表明，该算法明显优于其他六种算法。这些有希望的结果表明，AGSA比现有的多级图像阈值方法更合适。