In the research of the evaluation function, this paper firstly states the improvement Discrete Cosine Transform (DCT) algorithm and Small Univalve Segment Assimilating Nucleus (SUSAN) algorithm. Then, combining with the improved DCT algorithm and SUSAN algorithm, a new focusing evaluation algorithm has been presented. New algorithm collects the respective advantages of frequency domain algorithm and spatial domain algorithm, which enable the focus curve to has a greatly improvement in unimodal, local extreme points and sensitivity. In the research of the focusing window, a new selection method is proposed on the basis of the importance of image sub-block. The new method regards the gradient rate as image sub-block importance factor. Then, put the image sub-block whose factor value is less than threshold as background sub-block and removed. Finally, after removing background sub-block image, treat the rest of part as the focusing window. The new selection method can achieve dynamic distinguishing between target area and background area.