A performance evaluation of image interpolation and superresolution algorithms

Image interpolation and superresolution can be considered as a problem of estimating a high-resolution image from a single or multiple registered low-resolution images. In this paper we classify image interpolation and super-resolution estimators into linear and nonlinear, fixed and signal adaptive model, local and non-local, learning and non-learning-based, back-projection and non-back-projection. We compare the performance of different estimators for various gray and color images by using both objective measurements, such as MSSIM and PSNR, and subjective observation.