Digital image improvement by integrating images of different resolutions

Digital image improvement by integrating images of different resolutions

This proof of concept paper evaluates the performance of super-resolution (SR) imaging when combining a sequence of images of different resolutions. Traditional work in SR requires accurate sub-pixel registration and/or alignment techniques. The assumption is that all the frames in the sequence are captured with a random sub-pixel translation and at the same spatial resolution.
However, if there is no relative motion between the scene and the sensor, the super-resolution problem can be approached by integrating images of different resolutions using the image re-sampling ratios as the enhancing factor. This may be the case of using digital zooms whereby a scene of interest is enhanced by integrating the information obtained from capturing said scene at different zoom levels.
Preliminary results of conducting tests on synthetic and real data (scanned images) are presented. In both cases the process under-samples the image of an object of interest within a scene at different resolutions. By integrating these images using an algebraic process an improved composite is obtained containing more spatial information than that provided by simply interpolating on a single image.