针对现有图像分割算法存在分割效果不佳及耗时长等缺点，提出了一种二维 Otsu法红外目标阈值分割算法。根据局部邻域熵引入图像背景因子来对图像进行预分类，并采用最佳阈值与类内与类间方差的归一化处理对图像进行图像分类；最终通过约束阈值的搜索范围，来提高算法的准确性与稳定性。实验结果表明，该方法能够对不同类型的图像进行有效的图像分割同时保持较高的编码效率，与其他算法比较，该算法的图像分割效果及消耗时间具有一定的优势。

Considering the disadvantages of segmentation algorithms, such as poor segmentation effects and a time-consuming process, a two-dimensional Otsu infrared target threshold segmentation algorithm is presented. The image is pre-classified by introducing the background factor according to the local neighborhood entropy, and it is classified by the normalization of the best threshold and the intra-classand inter-class variances. Finally, the accuracy and stability of the algorithm are improved by restricting the search range of the threshold. Experimental results demonstrate that this method can effectively segment different types of images while maintaining high coding efficiency. Therefore, compared to other algorithms, our algorithm exhibits advantages in image segmentation and is less time-consuming.校级科研启动基金项目(2017YB005)