Reconstruction+Synthesis: A Hybrid Method for Multi-Frame Super-Resolution

In this paper a new method for multi-frame Super-Resolution is proposed. Having a low resolution(LR) video and a few high resolution(HR) images from a specific scene is frequently occurred. In addition it is usual that between images and video frames there is some differences, because of different of exposure time, moving objects, camera movement and so on. In the proposed method a Super-Resolution reconstruction method, applies on low-resolution frames for producing a reconstructed HR image. After that a synthesized image with mapping the high resolution training image to the reconstruction result, is produced using a proper transformation model. Fusion of the synthesized image with the reconstruction result makes the final desired HR image of the given LR frames. Experimental results show that our approach is competitive both for quality and quantity with some famous super-resolution methods.