The relationship between the feature point density and the object degree of focus is analyzed, and a focus measure based on the feature point density is proposed. A novel focus measure fusing the feature point density and the edge information is proposed. The focus measure for scene depth estimation and all-in-focus imaging from focus stack is utilized. The proposed method can effectively compensate for the focus measure based on the edge information which is inaccurate in the image texture regions. The focus measure fusing the feature point density function and the sum-modified-Laplacian (SML) is used for the three-dimensional scene reconstruction, and the scene depth estimation and all-in-focus imaging algorithm are achieved. The experimental results show that the proposed method can effectively eliminate the error depth value which the SML estimates in the image texture region, retain the advantages of SML in the edge region and the high precision scene depth maps and all-in-focus images are obtained.