为了改善自动对焦过程中爬山法的搜索性能和对图像清晰度进行更合理的评价,提出基于动态搜索步长的自动对焦算法和一种自适应图像窗口动态选择的图像清晰度评价方法。在用红外变焦镜头采集的多个变焦视频中提取图像,建立仿真模型比较了该方法和几种经典的对焦算法的性能差别,实验结果表明: 提出的对焦窗口搜索方法在视场中具有多个目标时,能进行更合理的评价；在进行对焦搜索过程中,本方法在过冲次数、最大过冲步数、行程步数和离焦程度这四个指标上具有明显优势,有效地减少了搜索过程中反复来回越过对焦位置产生的“拉风箱”现象,节约了搜索时间。

In order to improve the searching performance of hill-climbing method in auto-focusing process and evaluate the image definition more reasonably, an auto-focusing algorithm based on dynamic search step and an image definition evaluation method based on dynamic selection of real-time image window are proposed. In this paper, two consecutive zoom videos captured by infrared zoom lens are used to extract images, and simulation models are established to compare the performance differences between the proposed method and several classical focusing algorithms. The experimental results show that the proposed method can make more reasonable evaluation when there are multiple targets in the field of view. In the process of focus searching, this method has obvious advantages in the four indexes of overshoot number, maximum overshoot step, stroke step and defocus degree. It effectively reduces the phenomenon of "pulling bellows" which occurs in the process of searching, and save the focusing time.