Wide-band length-6 cubic interpolator

This paper proposes a weighted least-squares (WLS) method for designing wide-band length-6 cubic interpolation kernels in the frequency-domain. The length-6 cubic is symmetric and constructed by connecting three piecewise polynomials of third-degree, and their optimal coefficients are found through minimizing the weighted squared error between the desired and actual frequency responses of the cubic. The most significant feature of the WLS design method is that various cubics with different frequency responses can be easily designed through adjusting the weighting functions in different frequency bands, and “don´t care” bands can even be ignored. As a result, high-accuracy interpolators can be obtained for interpolating various signals containing different frequency components. A wide-band interpolation example is given to illustrate that the resulting length-6 cubic can achieve much higher accuracy interpolation (small interpolation errors) than the existing interpolators.