Super-resolution from multiframe X-ray images

X-ray dose reduction in certain occasions is particularly important because some of patients are very sensitive to radiation exposure. Meanwhile, diagnosis often requires zooming into a region of interest for more details in an image such as pediatric patients or Digital Subtraction Angiography. However because of hardware limitations of digital X-ray imaging systems, the zoom would contain a lot of pixel artifacts and insufficient detail. Super-resolution is an image processing technique that can produces images beyond the native resolution of a digital image sensor by accurate sub-pixel registering of aliased images , and offers large potential to fix the two problems above. In this paper, a super-resolution algorithm for Aliased X-Ray Images is proposed. In detail, firstly, we present a frequency domain technique to precisely register a set of digital X-ray images, which are shifted against each other by manually move the phantom,based on their low-frequency, aliasing-free part. Then a high-resolution image is reconstructed by using nonuniform interpolation. We demonstrate the ability of our approach to improve resolution while maintaining normal radiation dosages,and the resulting high-resolution image show good visual results, it also contains more detail, that can be exploited for zooming in diagnosis applications.