Evaluation of focus measures for the autofocus of line scan cameras

Plenty of focus measures have been proposed and researched in the past few years. However, little attention is paid to their performance in the autofocus of line scan cameras. This paper investigates sixteen typical focus measures, and their performance is compared for the autofocus of line scan cameras. A comprehensive evaluation method is proposed to assess the performance of focus measures. Subjective evaluation is firstly utilized to exclude the focus measures of poor performance. Then, objective evaluation which includes seven criteria is utilized to rank the rest focus measures. Scale factors are introduced for the evaluation of focus measures in different search algorithms and focus styles. Experimental results show that Tenengrad provides the best performance in both global-based search and local-based search in the autofocus with auxiliary focus sheet. In the autofocus directly on the object when the object is variable, autocorrelation provides the best performance in global-based search while Tenengrad provides the best performance in local-based search. In the autofocus directly on the invariable object, the best focus measure is related to image content, and it should be selected in field test.