基于图像矩函数的图像清晰度评价方法

自动调焦作为显微图像测量中的关键技术，其核心为图像清晰度评价函数。现有图像清晰度评价函数在对焦目标与背景的对比度较低时，评价曲线极易失去其理想特性，或灵敏度下降，造成聚焦精度的降低；或出现多个局部极值，造成调焦失败。针对这一问题，在分析离焦图像成像原理的基础上，提出一种新的基于图像矩的图像清晰度评价函数，并给出定义。新的评价函数由不同阶的图像矩加权构成，通过调节权重系数可以灵活控制评价函数的灵敏度和抗噪性。通过微小透明球体的显微图像序列对新的评价函数进行实验验证，并和四种常用评价函数进行分析比较。实验结果证明，在低对比度条件下，相对于其它四种常用评价函数，基于图像矩的图像清晰度评价函数在有效保持了如单峰性、无偏性等理想曲线特性的同时，在灵敏度及抗噪性方面也具有明显优势。

Image Definition Evaluation Function Based on Image Moments

Auto-focusing as the key technology of microscopic image measurement, its core was image definition evaluation function. When the contrast between focus target and background was relatively low, evaluation curves of existing image definition evaluation functions were inclined to lose their ideal characteristics with low sensitivity, causing the decrease of focusing accuracy, or multiple local extremums, resulting in the failure of focusing. Targeting at such problem, a new image definition evalua-tion function based on image moments was proposed on account of analyzing the image-forming principle of out-of-focus image, and its definition was presented. New evaluation function was composed by the weighting of image moments of different order, and the sensitivity and noise immunity of evaluation function could be flexibly controlled by adjusting weighting coefficient. Experi-mental verification on the new evaluation function was carried out by the microscopic image sequence of tiny transparent sphere. In addition, the analysis and comparison between the new evaluation function and four common evaluation functions were conducted. Experiment result showed that under low contrast, as compared to other four common evaluation functions, the image resolution evaluation function based on image moments not only effectively maintained ideal curve characteristics such as unimodality and unbiasedness, but also had obvious advantages in the aspects of sensitivity and noise immunity.