Photo-realistic continuous digital zooming for an asymmetrical dual camera system

Abstract

Asymmetrical dual camera systems are becoming more and more popular among many commercial products and they are expected to provide digital zooming images. Existing related algorithms such as super-solution and image fusion fail to reconstruct photo-realistic results because of the sampling rate inconsistency, especially when the focal lengths of cameras vary largely. Nevertheless, magnification factor of super-resolution algorithms are fixed and inflexible as well. By looking into the inherent drawbacks of existing super-resolution algorithm, a new photo-realistic continuous digital zooming algorithm is then proposed in this paper on the basis of a novel super-resolution framework and a band-inpainting algorithm. Experiment result illustrates that proposed algorithm combines the advantages of both super-resolution algorithm and image inpainting algorithm: it reconstructs band information with super-resolution methods, and introduce image inpainting algorithm to reconstruct poorly super-resolved texture areas to produce state-of-the-art photo-realistic digital zooming images.