Assessment of single-frame resolution enhancement algorithms

This paper provides subjective and objective a comparison of a multi-frame superresolution algorithm with four different single-frame algorithms. The multi-frame super-resolution reconstruction is based on an adaptive filtering approach, using least mean squares (LMS). The single-frame methods are, respectively, bicubic interpolation, wavelet-based interpolation, edge-adaptive interpolation and interpolation using wide sense Markov random fields. The performance is measured objectively with. PSNR and SSIM criteria, while the subjective evaluation is based on forced choice scoring. Wilcoxon rho test is employed to check fort e degree of agreement between subjective and objective results.