Structure Adaptive Super-resolution Reconstruction

Super-resolution(SR) image reconstruction deals with construction of a high-resolution(HR) image from a set of low-resolution(LR) images of a scene. It is one of the most spotlighted research areas as it can overcome the inherent resolution limitation of the imaging system and improve the performance in most digital image processing applications, such as remote sensing, medical imaging and security monitoring.All SR reconstruction algorithms can be divided into two categories, frequency domain method and spatial domain method. As spatial domain method has a natural advantage over frequency domain method in incorporating image prior information and using general noisy model, it has become the predominant method. Bayesian estimation is one of the most successful and promising methods among all SR reconstruction schemes, nevertheless it usually results in a smooth estimation with blurred edges. Enlightened by the rationale of bilateral filter, this paper employs intensity variation as prior information. The prior knowledge, which is actually the intensity variation extracted from the partially reconstructed image solved at the last step, is directly incorporated into the similarity term. As the adjusted similarity term can guide the cost function keeping out the outlier in the structure area, the proposed algorithm is structure adaptive and very successful in edge-preserving. Lots of experimental results show that the proposed algorithm has considerable improvement in terms of both objective measurements and visual effects.