Single-image Super-resolution using Complex Nonseparable Oversampled Lapped Transforms
Satoshi Nagayama, Shogo Muramatsu, Hiroyoshi Yamada · 2018
In this research, we propose a super-resolution method of complex image using Complex Nonseparable Oversampled Lapped Transforms (CNSOLTs). CNSOLTs are complex-coefficient transforms based on lattice structure. CNSOLTs are possible to design an analysis-synthesis system which simultaneously satisfies nonseparable, redundant, linear-phase, and compact support properties under the constraint of the Parseval tight frame. In addition, CNSOLTs are possible to select a redundancy with arbitrary rational number. The effectiveness of the proposed method is evaluated by super-resolution simulation of complex-coefficient images.