8th International Conference on Image Processing, Wavelet and Applications,, İstanbul, Turkey, 22 - 24 September 2016, pp.1-9
Resolution is a widely used term when judging various image acquisition/processing systems’ quality. In super resolution (SR) image reconstruction, complementary information from a set of different low resolution (LR) images are combined through aligning the images onto the same spatial coordinate system with sub-pixel accuracy. This opens a pathway for image resolution enhancement. Slightly different LR images of the same scene, which are accumulated closely, can be now geometrically refined using spatial codependency adaptations to construct a high resolution (HR) image. As a powerful geostatistical procedure, Kriging provides estimates of the values at unknown locations with a minimum quantifiable error and it is a completely data-driven approach. It is based on the measured codependency of data points, and provides a more accurate local representation of the data than polynomial or function based methods. We employ Kriging at the second stage of a standard SR image reconstruction method to increase the resolution of the images. Also a standard regularization procedure is accompanying the Kriging to provide a user-specified tradeoff between data fidelity and data smoothness. We applied the framework to synthetic and real world data and compared the results with those of several other interpolation methods, as well as state of the art image reconstruction methods.