Iterative-interpolation super-resolution (IISR)

Abstract

In the Oxford dictionary, the word interpolation has the meaning - "to insert (an intermediate term) into a series by estimating or calculating it from surrounding known values". In the digital age, image interpolation refers to the technique of recovering a continuous signal by estimating image data from a set of discrete image data samples. It links the continuous and the discrete domains. Image interpolation forms a fundamental base in image processing and is the heart of many computer vision applications such as medical imaging, target detection and recognition, and astronomical imaging. Almost every image processing software implements some interpolation technique for transformations, rotations and many other manipulations performed on an image. It is very important that the interpolation techniques have a very low computational cost in terms of both, time and memory utilization since they are usually implemented at some intermediate step in a system. At the same time, it is necessary for the technique to yield good and accurate results, or else it could jeopardize the final solution. For example, in the field of medical imaging, computed tomography (CT) or computed axial tomography (CAT) and magnetic resonance imaging (MRI) scan employ interpolation techniques during the registration process, a slight error in the interpolated data could cause mis-registration thereby significantly affecting the accuracy of reconstruction of the final image which may lead to wrong diagnosis of a patient. It is therefore very important to choose a correct type of interpolation technique, depending upon the nature of its application, which provides the best trade-off between accuracy and computational cost. © 2009 Springer-Verlag Berlin Heidelberg.