A Transform-Domain Approach to Super-resolution Mosaicing of Compressed Images

The combination of image mosaicing and super-resolution imaging, i.e. superresolution mosaicing, is a powerful means of representing all the information of multiple overlapping images to obtain a high resolution broad view of a scene. In most current image acquisition systems, images are routinely compressed prior to transmission and storage. In this paper, we present a robust super-resolution mosaicing algorithm which can be applied to compressed images. The algorithm operates on the quantized transform coefficients available in the compressed bitstream so that super-resolution reconstruction can be implemented directly in the transform domain. In order to improve the performance of super-resolution mosaicing, an adaptive approach to determining a regularization parameter is proposed. It is shown that this algorithm is robust against outliers and provides reconstructed super-resolution images with improved quality.