We present a vector angular spectrum approach using self-iterative algorithms for beam shaping. This approach is rigorous and valid for subwavelength feature size. It is based on scalar angular spectrum theory and deals with beam polarization along three components of the wave vector in the spectral domain during the iteration. All self-iterative algorithms can be kept efficiently for any incident polarization. Algorithms such as Gerchberg–Saxton, input–output, and phase mixture algorithm are used to design the diffractive optical element for super-Gaussian beam shaping in uniform illumination with traditional scalar diffraction integrals and this new approach, respectively. The comparison of top profile error, mean square error, and diffraction efficiency of the simulated output beams shows that this new improved approach is more efficient at keeping the variation of the beam polarization.