结核杆菌医学涂片大多具有观察区内容稀疏不均匀、杂质较多的特点, 使用自动显微镜检方法进行图像采集时, 会出现清晰度区分困难、效率低下、甚至聚焦评价失效的问题, 为提高自动镜检的效率和准确度, 本文自主搭建了显微视觉计算机自动检测系统, 对结核杆菌涂片的自动聚焦技术进行系统的研究。首先, 对比研究11种常用聚焦函数对结核杆菌镜检玻片图像聚焦评价的优劣, 并分析了聚焦成功和失效的原因。在综合分析各聚焦函数对结核杆菌涂片的聚焦效果基础上, 提出了一种基于Tenengrad的改进型聚焦评价函数, 通过改进内容像素的聚焦权重提高聚焦准确度, 优化图像处理算法来提高图像采集效率。实验结果表明: 改进型Tenengrad聚焦函数FTen-Q在结核杆菌涂片的各类视野图像评价方面具有高灵敏度和准确度, 其聚焦成功率和运算效率分别提高了13.884%和17.616%, 可以满足结核杆菌涂片类非均匀涂片的显微视觉自动检测应用要求。

Most of the medical smears of Mycobacterium tuberculosis have the characteristics of sparse and uneven content and more impurities in the observation area. There are some problems while observing the Mycobacterium tuberculosis smears to acquire image automatically, such as difficulty in distinguishing definition, low efficiency and focusing evaluation function invalid. To improve the efficiency and accuracy of automatic inspection, an automatic micro-vision detection system was developed independently to research the auto-focusing technology of the sputum smear images collection. Firstly, the sputum smear images focusing evaluation advantages and disadvantages of the eleven common focusing functions were studied comparatively, and the reasons for the image focusing failure were analyzed. According to the comprehensive performance of the various functions in the sputum smear images acquiring, an improved focusing evaluation function based on the Tenengrad focusing function was proposed. The image focusing accuracy was enhanced through adjusting the focusing weight of the content pixels, and the image processing algorithm was optimized to improve the image collection efficiency. The experimental results show that the improved Tenengrad focusing function (FTen-Q) has high sensitivity and accuracy in the images evaluation of Mycobacterium tuberculosis smear. Compared with the traditional Tenengrad function, the image focusing success rate and computing efficiency are improved by 13.884% and 17.616%, respectively, which can meet the application requirements of this kind of nonuniform smear micro-vision detecting system.