Multiframe Super-resolution Reconstruction Based on Information Fusion

Multiframe super-resolution reconstruction (SRR) can obtain higher resolution images regardless of limitations of physical sensors, which is widely applied in various fields. Most of the existing algorithms solve the under-determined SRR problem by adding prior information, of which the accuracy easily affects the quality of obtained high-resolution (HR) images. A SRR process through information fusion is proposed in this paper based on multiple low-resolution (LR) images with half-pixel displacements. HR images are generated by fusing the low-frequency estimation and extracted high-frequency information, and the optimal fusion factor are obtained resorting to one-dimensional search. The proposed process realizes more realistic image reconstruction due to the extraction of accurate HR gradient information instead of possibly inaccurate prior information and achieves excellent performance in objective quality. The simulation results show the proposed method achieved improvements of up to 9.2 dB, 18.1% and 7.9% in noise level, structure and feature indicators, respectively.