This paper proposes a method to reconstruct an all-in-focus (AIF) image using region-adaptive fusion from a series of limited depth-of-field images, called a focal stack. Based on the analysis of the focus profile, pixels are classified into three regions depending on the number of peaks. Different fusion rules are adaptively used according to the type of the region to construct the fusion decision map. Experimental results show that the proposed method reconstructs the AIF image with less artifacts than existing methods.