Image interpolation techniques are often required in electronic imaging for image generation, as well as for processing methods such as compression, warping, or re-sampling. Since the ideal interpolation function is spatially unlimited, several interpolation kernels of finite size have been published.

This paper introduces truncated and windowed sinc, nearest neighbor, linear, quadratic, and cubic polynomials, Gaussian and cubic B-spline interpolation, and approximation methods with kernel sizes from 1×1 up to 8×8. Various aspects of the techniques are compared, including spatial and Fourier analyses, computational complexity as well as runtime evaluation, and qualitative and quantitative interpolation error determinations for particular interpolation tasks.

The paper demonstrates that superior techniques are large-sized DC-constant interpolators.