Sinc interpolating approach to sub-pixel image stitching for virtual microscopy

Abstract

This study involves a two-step approach to the stitching of images taken for virtual microscopy application. The method has a sub-pixel resolution. To achieve this, first, a coarse shift level is adopted via a correlation function, and then a sinc interpolation is applied to the small neighborhood region of this coarse point. At the end, the sub-pixel shift level is determined by obtaining the maximum of the interpolated function using a gradient descent algorithm. Image stitching process is finalized with linear blending. Results are obtained within the range of ten percent subpixel tolerance.