In this paper, we present a novel approach to generate images of extended depth-of-field (DOF) to support realization of three-dimensional (3D) imaging systems such as integral imaging. In our approach in extending the DOF, we take advantage of the spatial frequency spectrum of the object specific to the task in hand. The pupil function is thus engineered in such a fashion that the modulation transfer function (MTF) is maximized only in these selected spatial frequencies. We extract these high energy spatial frequencies using the principal component analysis (PCA) method. Moreover, given the need for many pupil function engineering steps in 3D imaging systems, we have constructed an approximate expression for MTF to be used in the design of optimum amplitude and/or phase pupil filter. Moreover, we have optimized the DOF extension process with blocking the minimum possible area in the pupil plane. This maximizes the output image quality (e.g. 10% DOF improvement) compared to the existing methods where non-optimal blocking of the lens area may cause more degradation in output image quality. Experimental results are presented to illustrate our proposed approach. © 2010 IEEE.