光学技术, 2019, 45 (2): 218, 网络出版: 2019-04-28
角膜曲率测量中自动识别的对焦评价
Focus evaluation for automatic recognition in corneal curvature measurement
图像处理 角膜图像 对焦评价 对数灰度级 四方向Sobel算子 image processing corneal image focus measures logarithmic grayscale four-direction Sobel operator
摘要
角膜曲率测量一般是采用视标成像的方法得到角膜的反射像后, 通过对反射像的再成像间接得到角膜的曲率半径。为了实现对角膜曲率测量图像的对焦判断, 在简化角膜曲率测量系统的基础上, 对获取的图像对焦算法进行研究。对角膜曲率测量图像的六个测试光标分别进行像素统计, 初步判别过度离焦的状态。对角膜曲率测量图像进行四方向Sobel算子的卷积运算, 对得到的边缘增强图像进行对数灰度级像素统计, 得到对焦评价值。将该算法与Tenengrad函数、方差函数和单一对数灰度级像素统计评价函数算法做对比实验, 评价该算法的可行性。实验结果表明, 该对焦评价函数具有良好的精确度、单峰性、灵敏度和无偏性, 对细节信息较少、结构简单的图像的对焦评价方面有应用前景。
Abstract
The corneal curvature measurement is generally obtained by using the method of visual imaging to obtain the reflection image of the cornea. The corneal curvature radius is indirectly obtained by re-imaging the reflected image. In order to achieve the focus judgment of the corneal curvature measurement image, the acquired image-focusing algorithm is researched based on the simplified corneal curvature measurement system. The six test cursors of the corneal curvature measurement image were separately counted in pixels, and excessive defocusing state was judged and eliminated initially. Gradient image was constructed by using four-direction Sobel operator with corneal curvature measurement image. The acquired edge enhanced image was used pixel statistical based on different gray level. In addition, focus evaluation value was obtained. The algorithm was compared with the Tenengrad function, the variance function and the single logarithmic gray level pixel statistical evaluation function algorithm to prove the feasibility of the algorithm. The experimental results show that the proposed focus evaluation function has good accuracy unimodality, sensitivity and unbiasedness. Therefore, the proposed method has a potential application for image focusing evaluation with less detail and simple structure.
方丽萍, 陈兆学, 陈鹏, 王成, 张大伟. 角膜曲率测量中自动识别的对焦评价[J]. 光学技术, 2019, 45(2): 218. FANG Liping, CHEN Zhaoxue, CHEN Peng, WANG Cheng, ZHANG Dawei. Focus evaluation for automatic recognition in corneal curvature measurement[J]. Optical Technique, 2019, 45(2): 218.