中文摘要：

利用经典的Otsu算法和基本遗传算法相结合进行图像分割存在有算法效率低、容易提前形成伪解的问题，对于上述问题，提出一种基于改进小生境遗传算法的图像分割算法（IVNGAMS）。算法全局优化了二维Otsu图像分割函数，可以按照个体适应度大小自动控制遗传参数。并通过引入模拟退火算法，进一步提升算法的局部搜索能力。实验结果表明，改进的图像分割方法能更好提升算法的全局搜索能力，能够更加稳定快速的收敛到最佳的分割阈值，并且得到了更好的图像分割效果。

英文摘要：

Aiming at the problem of the influence of the same query keyword and different binarization algorithms on the overall retrieval performance in historical Mongolian document images retrieval, this paper presents an image binarization method of historical Mongolian document based on Markov random field to improve the retrieval performance of historical Mongolian documents. The MRF model is used to model the gray level image and the binary image. The prior probability of the hidden-layer is estimated by the training codebook, and the probability density of the observable-layer is estimated by analyzing the histogram of the gray image. The two kinds of prior knowledge are used to realize image binarization. The experimental data set is 100-page Mongolian Kanjur. In order to verify the performance of the proposed method, R-Precision is used as the evaluation index. Experimental results show that the binarization method based on Markov random field can not only effectively repair the damaged image, but also can improve its retrieval performance.