提出一种采用图像直方图特征(Histogram Feature, HF)函数的自动曝光方法, 用于在背景光照快速、大范围变化的情况下对高速相机进行自动曝光控制。首先, 采用多点测光对获取的图像进行兴趣区域(Region of Interests, ROIs)提取, 以降低系统测光的计算量。然后, 通过计算兴趣区域的HF函数选取大步长对曝光时间进行粗调。最后, 通过模糊逻辑计算出曝光时间的精调步长并运用阈值限制实现变步长搜索最佳曝光时间, 提高高速相机自动曝光的准确性及稳定性。实验结果表明: 本文方法可以在2 ms内完成一帧图像的亮度测量并对曝光时间进行调整, 相较于基于平均亮度值的传统自动曝光方法, 在0~110 ms内, 光源光照强度从760~23 100 lux反复变化时, 本文方法获得的图像信息熵比平均亮度值法的信息熵均值提高48.38%, 方差降低62.13%, 可以提供更好与更稳定的图片细节信息, 为后续的自动调焦、图像识别以及目标跟踪提供参考。

In this paper, a method for auto exposure based on image HF (Histogram Feature) function was proposed to achieve auto exposure control on high-speed camera under the condition where the background lumination was in fast speed and wide-range changes. First, multi-point photometry was conducted to extract the Regions Of Interest (ROI), thus lowering calculation amount of the photometric system; then, large step size was selected to carry out coarse adjustment of the exposure time by calculating HF function of these ROIs; finally, accurately-adjusted step size of the exposure time was calculated through fuzzy logic and then the optimal exposure time was achieved through variable-step search by using threshold limit, so accuracy and stability of the high-speed camera auto exposure were improved. The resuts show that proposed method is able to complete brightness measurement of one frame image as well as an adjustment in the exposure time within 2 ms; and compared with traditional auto exposure method based on average brightness, average information entropy achieved by proposed method is improved by 48.38% within 0-110 ms when the illumination intensity varies repeatedly from 760 lux to 23 100 lux, with a variance decreased by 62.13%. This method can provide better and more stable details on the image, which serves as useful reference for follow-up auto focusing, image recognition and target tracking.