Super-Resolution algorithm with context switching motion estimator and POCS-based reconstruction core

In the paper Super-Resolution algorithm with enhanced reconstruction stage and the context switched motion estimation is presented. The switching motion estimation concept stems from the simple premise that Fourier analysis method may be not accurate when the spectrum is destroyed by aliasing. In such situations Taylor series expansion is favorable. In the frequency domain method the properties of relocated images Fourier transforms are used to estimating of rotation and translation, as well as subsampling artifacts. In this paper I propose a concept based on switching motion estimator between spectrum and Taylor series methods. The choice of one of them depends on an aliasing factor, which I can evaluate. Previously, in the reconstruction phase the bicubic interpolation has been usually employed, in this paper it is replaced by iterative projection onto convex sets with biharmonic spline interpolation approach at its core. The new algorithm indeed gives improved results if compared to the previous ones.