基于图像熵的自动聚焦函数研究

Research on auto-focused function based on the image entropy

从图像中熵的定义出来,研究了图像熵在成像系统自动聚焦判决函数中的应用,研究发现:图像熵作为聚焦判决函数出现了两类不同的判决准则,一类是图像清晰熵小,另一类是图像清晰熵大。对不同类型测试图像的分析和计算,证实了这种非一致性的存在并分析了造成这种现象的可能原因,而且由于图像熵的峰值较小,数值仅在小数点后变化,不易找出全局最大值。在均方差判决函数的基础上,提出了条件加权均方的方法,得到了具有准则一致性的自动聚焦判别函数。实验证明:条件加权均方差判决函数在不增加运算量的前提下,可以加大函数的峰值,从而提高判决的准确性。实用中在测试条件变化不大的情况下,通过人机交互,可选用图像熵作为自动聚焦判决函数,否则优先考虑条件加权均方函数。

Based on the definition of entropy in image, the application of the image entropy in auto-focused function is studied. It is found that there are two different rules in focusing function while using the image entropy. When the image is in the state of focus, one with the entropy is the largest, and the other with the entropy is the least. Through calculating and analyzing different kinds of test images, the existence of rule variance is confirmed, and the reason is analyzed. As the peak value and its shift are small, it is difficulty to find the largest in the four corners. Based on mean-square judge-function, a method weighted in condition is given, and an auto-focus judge-function with coherence rule is obtained. The experiment proves that the mean-square judge-function weighted in condition can enhance the peak of function and does not increase amount of calculation, so it can improve the veracity of judgement. If the test-condition does not change a lot, image entropy can be acted as an auto-focus judge-function through human-computer interface, otherwise the conditional weighted mean-square function is recommended.