On Auto Focusing under a Microscopic View

Auto focusing (AF) is an important technique in applications for precision measurement and inspection. The technique is currently developed more focusing on high-speed implementation and under a microscopic view. This paper presents an AF algorithm composed of three stages (i.e., initial search for direction, a rough search, and a fine search). Sophisticated decision makings and abnormality handlings are built in to enhance the performance of the proposed algorithm. It is illustrated in this study that previous measures of image focus value (FV) are subject to low signal-to-noise ratio. Hence, a new measure based on the variance of sub-windowing measures is proposed to enhance the ratio. By the enhancement, the AF reliability and accuracy can be effectively improved. The proposed AF algorithm has been implemented in a PC- based laser-repairing machine. Experimental tests have been widely conducted in lab and on line as well. Selected tests are included in this paper. Among the tests, the AF time spreads over the range of 400-1700 ms with the average AF time less than 1000 ms.