A fast autofocusing algorithm based on prediction is proposed for the large-scale microscopic images acquisition of flat preparations. Firstly,the method uses the in-focus positions of adjacent images to estimate that of current images, and then makes tiny adjustment at the estimated position to get accurate in-focus position,thus,avoids the local maxima problem of focusing curves. Secondly, the global threshold method is used to select focusing window adaptively, and it can eliminate the background interference and ensure the accuracy of focusing curves. This algorithm decreases greatly the number of images that are acquired for autofocusing,and increases the speed,accuracy and robustness of the microscopic autofocusing sys tem.