A pixel based multi-focus image fusion method

Abstract

The lenses used in computer vision systems by nature provide limited depth of field, so digital cameras are incapable of acquiring an all-in-focus image of objects at varying distances in a scene. To obtain an all-in-focus image of such scenes from their differently focused images, this paper proposes a novel spatial domain multi focus image fusion technique. The developed technique, firstly, computes the point spread functions (PSF) of the source images. Next, the images are artificially blurred by convolving them with the estimated PSFs. Then, the artificially blurred images are used to determine the sharpest pixels of the source images. Finally, the all-in-focus image of the scene is constructed by gathering the sharpest pixels of the source images. Experiments are carried out on several multi-focus image sets. The proposed method and other well-known image fusion methods are compared in terms of visual and quantitative evaluation. The results obtained show the feasibility of the developed technique.