For the case of atomic force microscope (AFM) automation, we extract the most valuable sub-region of a given AFM image automatically for succeeding scanning to get the higher resolution of interesting region. Two objective functions are sum-marized based on the analysis of evaluation of the information of a sub-region, and corresponding algorithm principles based on standard deviation and Discrete Cosine Transform (DCT) compression are determined from math. Algorithm realizations are analyzed and two select patterns of sub-region: fixed grid mode and sub-region walk mode are compared. To speed up the algorithm of DCT compression which is too slow to practical applied, a new algorithm is proposed based on analysis of DCT's block computing feature, and it can perform hundreds times faster than original. Implementation result of the algorithms proves that this technology can be applied to the AFM automatic operation. Finally the difference between the two objective functions is discussed with detail computations.